Article 9 of vmsnet.decus.journal: Path: nntpd.lkg.dec.com!crl.dec.com!crl.dec.com!bloom-beacon.mit.edu!apollo.hp.com!lf.hp.com!news.dtc.hp.com!col.hp.com!simtel!news.sprintlink.net!news.clark.net!rahul.net!a2i!bug.rahul.net!a2i!news.erinet.com!uunet!in1.uu.net!noc.near.net!news3.near.net!eisner!mcmahon_b From: mcmahon_b@eisner.decus.org (Brian McMahon, Info-VAX Refugee) Newsgroups: comp.org.decus,vmsnet.decus.journal Subject: DECUServe Journal July 1995 Message-ID: <1995Jul27.084840.10695@eisner> Date: 27 Jul 95 08:48:39 -0400 Organization: DECUServe Approved: mcmahon_b@eisner.decus.org Lines: 3092 Xref: nntpd.lkg.dec.com comp.org.decus:7330 vmsnet.decus.journal:9 The DECUServe Journal --------------------- July, 1995 From the Editors' Keyboard . . . . . . . . . . . . . 2 What's Inside, Changed Format, WWW Stuff, et cetera WWW File Serving . . . . . . . . . . . . . . . . . . 3 When Anonymous FTP isn't an Option Identifying Web Clients . . . . . . . . . . . . . . 5 Which Client? Which Version? "Standard"? Mosaic Display Problems . . . . . . . . . . . . . . 8 WPS-PLUS, ALL-IN-1, and WWW PC Motherboard Replacement . . . . . . . . . . . . 13 Tips for Upgrading a PC Testing for Removable Disk . . . . . . . . . . . . 16 Using IO$PACKACK with Removable Media Force Logging of a Session . . . . . . . . . . . . 19 Keeping Logs of a Privileged Account SYSUAF in Mixed VMSclusters . . . . . . . . . . . 22 User Quotas on VAX vs. Alpha MOUNT and OpenVMS V6.1 . . . . . . . . . . . . . . 26 Changed Security Behavior Windows NT Crash Dumps . . . . . . . . . . . . . . 29 Tips and Tricks for NT Crash Dump Analysis SET SECURITY on Tape Units . . . . . . . . . . . . 32 Persistence of the SET SECURITY Command UNIX Filename Globbing . . . . . . . . . . . . . . 33 C-Kermit Strikes Again POP Server for Unix . . . . . . . . . . . . . . . 35 Where to Get a POP Server 4mm DAT Autoloader . . . . . . . . . . . . . . . . 37 SCSI Commands to Control TLZ06L Autoloader RRD50/KRQ50 Info . . . . . . . . . . . . . . . . . 43 Hardware Details for Old CD-ROM Wildcard Algorithm . . . . . . . . . . . . . . . . 44 Pattern-matching with code in C DTR column headers . . . . . . . . . . . . . . . . 47 Controlling the printing of column headers in Datatrieve Saving Window Positions . . . . . . . . . . . . . 50 Tips on DECwindows Settings The DECUServe Journal July, 1995 Page 2 From the Editors' Keyboard From the Editors' Keyboard ---- --- -------- -------- Once again we ask you to pardon the dust and construction debris. We're changing the format of the Journal somewhat, to make it more suitable for HTML and the Web. At the same time, we're trying to keep things simple at the production end of things, improve readability, and still keep a format that is suitable for plain-text transmission and even good old-fashioned hard copy printing. You've probably not seen the last of the format changes. Speaking of the Web, several articles in this issue are taken from the new WWW Notes conference on DECUServe. There, topics range from server security to client quirks to interesting URLs. Needless to say, it's a fairly lively place. But this issue doesn't focus exclusively on the Web, so if you're sick of hearing about it, don't despair. In fact, this issue of the Journal is rather more eclectic than most, even by our own standards. We've collected a broad sampling of DECUServe's treasures, mostly of recent shorter pieces this time, and some from seldom-visited conferences -- ALGORITHMS, for example. We hope you'll find the results useful and enjoyable. No matter whether you love it or hate it, we always welcome your feedback. Editorial contact information appears in every issue. Let us know what you think, about form as well as content. * * * * * The DECUServe Journal July, 1995 Page 3 WWW File Serving WWW File Serving --- ---- ------- Abstract: Consider this thorny problem: You want to do file service in HTTP, but for one reason or another, the usual anonymous FTP setup is not workable. Can you do it at all, and can you do it without blowing a great gaping hole in your site's security? The situation is a bit convoluted, to be sure. [Administrative Note: The WWW conference has adopted the convention of prefixing note titles with a code to identify the kind of topic for ease of browsing. Several codes have been defined, such as URL for interesting spots on the Web, HTTPD for server-related concerns, and UI for user-interface (Web browser) issues. We've decided, at least for now, to retain these prefixes in the Journal.] Participants: Dan Esbensen, J. M. Ivler, Terry Kennedy, Harrison Spain, Bret Wortman. Conference: WWW Note 39.0, 14-Jun-1995 Spain: HTTPD: Using HTTP to server files w/o Anon FTP ----------------------------------------------------- We are struggling to implement file serving (both send and receive if possible) using our HTTP server. Unfortunately, security restricts the use of Anonymous FTP! Each user of the system has a unique username and password. Is there a way to serve files using an HTTP server *without* Anonymous FTP? We found a way to specify the username and password but if you access a document this way the file is displayed to the screen (yikes) as binary (argh). What we need is the same mechanism as Anonymous FTP but *with* a username and password. NOTE: We are running the Ohio State University HTTP server on our AXP/VMS system. Any pointers will be *greatly* appreciated! :-) Note 39.1, 14-Jun-1995 Esbensen: ftp://username:password@xxx.yyy.zzz/ ?? -------------------------------------------------- Does: ftp://username:password@xxx.yyy.zzz/ The DECUServe Journal July, 1995 Page 4 WWW File Serving work for you? Note 39.2, 14-Jun-1995 Ivler: Look at mime.types or srm.comf (if you have them) -------------------------------------------------------- Putting that on a page (I assume you mean to hardcode a username:password combination) would be deadly in terms of security. I think Harrison wants to modify the mime.types file on his server to accept certain extensions as "binary" transfers. This would allow him to use the http://bbs.ug.eds.com/filename.special_ext format, yet have the browser accept is as a download binary without resorting to anon FTP. Note 39.3, 15-Jun-1995 Spain: Getting closer! ---------------------- > ftp://username:password@xxx.yyy.zzz/ This works, but it interprets the file format differently than Anon FTP. For example, if you try to download a notebook (with extension _.NOTE), Netscape will try to display the file :-). I think JM hit the two concerns. Concern #1: As long as the password does not display on the screen, I think we are OK. Concern #2: We do have to figure out how Anon FTP figures out the file type. One alternative is to rename all the file to a couple of std extensions (one for text and the other for binary) but this would be a tad messy :-). Note 39.4, 15-Jun-1995 Kennedy: Ummm... ---------------- > Concern #1: As long as the password does not display on the screen, I > think we are OK. What about "View source"? The DECUServe Journal July, 1995 Page 5 WWW File Serving Note 39.5, 16-Jun-1995 Wortman: One idea.... --------------------- How about defining a form, allowing each accessor to enter username/password combinations, and using these during in the ftp://... string? Haven't tried this myself, just brainstorming. Note 39.6, 19-Jun-1995 Spain: I suspect this can't be done using HTTP ---------------------------------------------- We do have a partial solution. Since we validate our users before they are allowed access to this area, we already have the username and password. As long as we name all the files with either .exe or some special extension the ftp://... format will work. We can't get the file directory to list anything outside of the user's SYS$LOGIN directory though :-}. Note 39.7, 21-Jun-1995 Ivler: Show and tell works better than tell - sometimes ------------------------------------------------------- Harrison, since I was over there anyway, I dropped in to see Barry and walked him through the process I described. The proposed solution is what I'm posting. A page will be created after the user enters their name/password combination. This will offer them the ability to go to a download area. If they select that, it will go to a seperate WWW server on the same machine (ex: http://www.ug.com:8080/). That macine will be running with "index" on. Passing the server name without the file will create an indexed list of files for all files in that default area (your download directory will be the WWW_ROOT:). The mimes.types will be set to alolow .txt, .html and readme.* to be the only files to be text/* mime typed. All others will be binary. This will allow the http: to download these files as binaries onto the users machine. It totally avoids the FTP issue. Identifying Web Clients ----------- --- ------- The DECUServe Journal July, 1995 Page 6 Identifying Web Clients Abstract: The wonderful thing about standards, they say, is that there are so many to choose from. Or, in an even worse case, so many distinct variations on the same supposed standard. This pretty well describes the current state of HTTP. So it's natural to want to identify, as well as possible, what kind of browser (with which capabilities) has connected to the server. It's also natural to wonder whether this isn't precisely the sort of mess that Internet standards are supposed to avoid. This question, shall we say, exceeds the scope of the Journal (and the DECUServe Canons of Conduct). Participants: Bob Hassinger, Benjamin Pena, Kevin Roels, Pat Scopelliti. Conference: WWW Note 36.0, 8-Jun-1995 Scopelliti: HTTPD: does server know client type? ------------------------------------------------ Is there any way for the HTTPD server (or a script) to determine the type of browser that the client is using? I like using , but LYNX and Mosaic (at least the versions we currently have) do not support it (Netscape V1.1 does). So, I'd like to be able to determine if I can use
...
or instead have to use 
...
(yuck). Oh, this would be used by a script so it could output the right HTML. I found one hack.. appending
at the end of each table row at least prevents LYNX and Mosaic from wrapping everything, but Netscape seems to leave one blank line above the table for each row in the table. Note 36.1, 9-Jun-1995 Roels: Version Too? ------------------- It gets worse. From what I understand from our webmeister, the latest beta of Mosaic supports tables. I know you also wanted the job of determining version........ The DECUServe Journal July, 1995 Page 7 Identifying Web Clients Note 36.2, 9-Jun-1995 Hassinger: Yes. Request Header User-Agent ------------------------------------------ The latest version of Mosaic for the Mac that I have found is V2.0.0 beta 1 3/6/95. It does "support" tables. However, tables I have designed that look good viewed with Netscape 1.1 do not look as good in Mosaic. One of the Request Headers that browsers send is User-Agent. As an example, using the NCSA Mosaic browser for VMS and the OSU server, in the server script I get a value for this header of "NCSA Mosaic for the X Window System/2.4A libwww/2.12 modified". This seems to be simply a string with limited standardization of the format since each browser can send what it wants. That is, its not really clean, structured data. I suspect you would need to find out what the specific browsers you are interested in send and look for it explicitly. In general the format seems to just be specified by HTTP as "program/version comments". I am not certain if this header is required to be sent, but I think you can depend on a particular browser you are looking for to send whatever it sends every time. So the logic would need to be, use 
...
_unless_ you see a browser you know, in which case you would use ...
, perhaps even with further variations for particular browsers and versions if you are pushing the feature envelope. As HTML V3 develops and we get an increasing range of browser capabilities out there I rather expect we will see more and more of these issues. Note 36.3, 9-Jun-1995 Roels: String Value ------------------- I believe the string is http_user_agent, the server returns whatever the developer of the client puts in there. It really isn't the most consistent data you ever saw, we have registered 21 different client browsers hitting our site. Note 36.4, 9-Jun-1995 Hassinger: Details vary by context. And HTML V2 to V3 transition... -------------------------------------------------------------------- Judging from the information I have, the HTTP standard refers to the header as "User-Agent". In the case of the OSU server, in the DCL script case, you find it in a symbol named WWW_HTTP_SERVER_AGENT. The details of how this information is passed appears to vary from one server to another. But, in any case, the answer to the question is yes, you can tell something about the client browser you are communicating with, but it The DECUServe Journal July, 1995 Page 8 Identifying Web Clients is not neat. I think it also multiplies the effort when you start trying to accommodate different servers with special case code. BTW - I noticed in the HTTP 3.0 information a "recommendation" that V3 documents be transmitted with a MIME content type of "text/html; version=3.0". The idea is that existing HTML V2.0 user agents will be caused to safely handle them by displaying a save to file dialog rather than trying to incorrectly display the document as if it was HTML V2.0. This sort of thing will, to some degree, further complicate this subject. Note 36.5, 9-Jun-1995 Pena: point of information -------------------------- Point of information: where exactly is the 'official' documentation for the standard as it evolves and the finalized standard ? I've seen some information on HTML 3.0 that were from December '94, and didn't seem to have been updated all that often since then (and also was a draft) so I'm not certain if: 1) HTML 3 was ever finalized to date 2) there's a better site for information on HTML standards Note 36.6, 9-Jun-1995 Hassinger: http://www.w3.org/ ?? --------------------------------- I can not say with any authority at all because I am not close enough to the process. However I have noticed the material posted at http://www.w3.org/ looks like it may be a good starting point. Mosaic Display Problems ------ ------- -------- Abstract: Eh, did he say ALL-IN-1 and the Web? Indeed he did! (There was even a session about it at DECUS '95/Washington D.C.) This discussion began as a problem with a few documents converted from WPS-PLUS and not displaying properly in Mosaic. It quickly took on a much more far-reaching scope. The DECUServe Journal July, 1995 Page 9 Mosaic Display Problems Participants: Bob Hassinger, J. M. Ivler, Lynda Peach, Graham Pye, Pat Scopelliti, Don Vickers. Conference: WWW Note 40.0, 20-Jun-1995 Peach: UI: ASCII files display differently -- why? -------------------------------------------------- Why is MOSIAC displaying some files correctly and others with a totally different font and all tabs, paragraphs ignored. For example, if these two paragraphs were displayed by the 'bad font' the paragraph would be ignored completely and the font size would be much bigger. The files are in the same directory, placed there by the same method and are all converted to ASCII. The original files were WPS-PLUS most of the time. But two WPS-PLUS files can be converted to the directory and yet one is readable, one isn't. This is an Alpha, VMS version 6.1. Note 40.1, 20-Jun-1995 Vickers: Any document structure at all? --------------------------------------- It's not clear to me whether the WPS-PLUS files are being converted to 'just' ASCII or whether there are HTML tags being added as well. You might wish to describe the document structure a bit more. It might help if you post a couple of small examples of the documents in ASCII. Note 40.2, 20-Jun-1995 Peach: More details on document structure, conversion, etc. ----------------------------------------------------------- These are ALL-IN-1 files -- some may be WPS-PLUS files and some may be generated with EDT or EVE editor. No HTML tabs are involved. A command procedure and a couple of ALL-IN-1 scripts copies the files from their original set of ALL-IN-1 drawers into ASCII files in various VMS directories. The actual script code that does this is (for ALL-IN-1'ers this will have meaning): COPY #RealDoc #Fileconv_Dest_File TXT$DSAB_ASCII The file names are the titles given the documents in ALL-IN-1 with invalid characters eliminated or replaced by _'s. Because of the The DECUServe Journal July, 1995 Page 10 Mosaic Display Problems character limit in file names, this means files do not have a .TXT extension. For example: $$14HOSPITAL_ACCESSIBILITY_PIT.;1 $$16FACILITIES_PROGRAMMING_AND_DESIGN_P.IT;1 If one does a DIR/FULL on a "good" file and a "bad" file, you get the same on each: Record format: Variable length, maximum 74 bytes <-- byte number Record attributes: Carriage return carriage control changes We have even tried copying the "bad" files to short names such as TEST.TXT -- no effect. +---------------+ | Original File | | in ALL-IN-1 +------+ +-------+-------+ | +========|==============|===+ | Convert w/ | | COPY ... TXT$DSAB_ASCII | +-------------+ +========|==============|===+ | NCSA | +------+ | | MOSAIC | +---------------- xxx.xxxx | | | | OK | | +---------+-+---------+ | | +---------+ | | | | | | +---------+-+------------+ +--------+ | +---------+ | +-------- yyyyyy.yyy +-------------+ BAD The one thing we have found out is if we edit a "bad" file and put even a few characters such as XXX in... Then the display of that file becomes good. Any help would be greatly appreciated. Note 40.3, 20-Jun-1995 Vickers: Try 
 and
tags ---------------------------------- It sounds like Mosaic is processing the 'bad' files as if they are HTML files and the good ones as if they are plain ASCII. HTML files require paragraph markers (

), for example. The external file attributes probably have little to nothing to do with the treatment. It MAY be that the file extension would have SOME effect but I am not sure. I suggest that you may wish to enclose the text in your files inside of 

 and
tags to indicate to the browser that the file is pre-formatted. Place 
 as the first line of the file and
as The DECUServe Journal July, 1995 Page 11 Mosaic Display Problems the last of one your 'bad' files to see if this changes the behavior. Adding these tags would be very easy, of course. Note 40.4, 20-Jun-1995 Scopelliti: Special characters? ------------------------------- One other thing to check: The characters "<", ">" and "&" have special meaning. Does the bad document contain some of these? If so, Mosaic may be fooled into processing some of the text as HTML markers leading to undesirable results. Note 40.5, 20-Jun-1995 Ivler: Another method --------------------- also you can force it to use certain types by the extension. look in your servers mime.types file and see what the extention default is for ascii text (it shoud be txt) and set the filename to that default extention. Your server should then server the document with the default mime type, and the browser should know how to parse the file. Note 40.6, 21-Jun-1995 Pye: Coming soon to a system near you ------------------------------------- I guess you'll be very pleased when our ALL-IN-1 WWW interface is available in V3.2... Note 40.7, 21-Jun-1995 Hassinger: OH! More info please. Maybe there is still hope... --------------------------------------------------------------- I hope you are serious. I am currently desperately looking for a way to justify continuing to pay the support costs for ALL-IN-1 on my system. It seems to have become largely irrelevant to the goals of our management. OTOH, WWW is a VERY hot topic... Could you give some sort of broad overview of what you are thinking of? Would ALL-IN-1, say running on a VMS system present its menus and applications via HTTP/HTML to standard web browsers? Would it need server software or do the job itself? Would this capability be included in the normal updates or would it be a new cost? What time frame (like "this calendar year" for example). That sort of thing... The DECUServe Journal July, 1995 Page 12 Mosaic Display Problems Note 40.8, 21-Jun-1995 Hassinger: Some thoughts and questions - HTML takes new ways of thinking... --------------------------------------------------------------------------- As to the original question - I am still trying to get a handle on the exact symptom to help narrow the problem down to the server side or the client/browser side. A lot of the problems I have seen derive from having a lot of different servers out there, each somewhat different and every copy configured differently, and lots of different clients, and even different versions of each of the clients out there. HTTP/HTML seems much more vulnerable than much of what we are accustomed to working with. With so many mix and match combinations and the shift of a lot of the control over presentation format from the host to the client, its a new world in many ways for old timers... Can you describe in more detail the change in fonts you are seeing? Is one of them the one your browser uses for 
 presentation -
typically something like a mono-spaced typewriter font such as Courier? 
Or is it just a different size of the normal presentation fonts?  Can
you look in your browser setups and see if you can identify the
particular fonts and sizes as being associated with particular
configuration settings (e.g. for certain size headings, etc.).

Many browsers let you configure these things so your copy and mine of
the same browser might give different presentations because of the way
we have them set up.

In fact, did you mention what server and what client(s) you are seeing
this with?  And, have you compared the results using more than one
client?  For example, I try to test using the latest versions of both
Netscape and Mosaic on the the Mac, as well as Mosaic 2.4 for VMS, and
when it makes sense, Lynx on VMS too.



Note 40.9, 21-Jun-1995
Peach: One answer -- and I will get the rest.
---------------------------------------------
Lots of good questions and suggestions... will be looking into each and
everyone and will get back to you.
    
Re Ivler's suggestion about the file extension.  We cannot change the
file extension because we need *ALL* the characters available in both
the filename and extension as a way of providing the title.  Now, I
have thought of another way to do it if that becomes absolutely
necessary but since we have been able to prove that if the file is
unreadable regardless of whether we copy it to something.TXT, I
download it via my script to something.TXT -- it just doesn't matter
what the extension is or isn't.  The display is screwed up on some
files.
    
Now -- I'm going to extract this note and try to get as many answers






























The DECUServe Journal  July, 1995                              Page 13
Mosaic Display Problems


online as quick as I can.
    
P.S. to Graham -- Is the ALL-IN-1 and WWW *part* of 3.2?  Or will it be
offered like Personal Assistant?  With an addiitional fee?  If that is
the thought, please convey to the powers that be that they need to
rethink and quickly.  It is getting hard and harder to justify the
support.  Providing this as part of 3.2 could very well be the handle
that many of us desparately need
    


Note 40.10, 27-Jun-1995
Pye: More on the ALL-IN-1 / WWW interface
-----------------------------------------
Re .7
    
It's not a full interface to all of the functionality. It's a file
cabinet browser. But it needs the 'information provider' on the
ALL-IN-1 system to do nothing. The WWW interface code dynamicallys
builds HTML from the file cabinet. In demos, we show VT, TeamLinks and
WWW screens side by side, and as you create a document from VT or TL,
the web browser can magically see it!
    
Re .9
    
Here's the scoop. You'll need some code to run on your WWW box, which
dynamically builds the HTML and talks to the File Cabinet Server. This
will be free. We haven't completely decided which platforms this will
be available on yet - Votes?
    
Then you'll need the File Cab. Server DSO licence, either your existing
one for FCS-FCS access or a cheaper variant for WWW access only.




  PC Motherboard Replacement
  -- ----------- -----------

  Abstract:

  An "in-cab" upgrade is a familiar concept for many a grizzled old
  VAX manager; but how about for that long-in-the-tooth Intel PC over
  in the corner?  Here are some things to keep in mind when
  considering a motherboard swap in a PC.



  Participants:

  Charlie Byrne, Pierre Hahn, Jamie Hanrahan, Terry Kennedy, Jeff
  Killeen.
































The DECUServe Journal  July, 1995                              Page 14
PC Motherboard Replacement


  Conference:  PERSONAL_COMPUTING


Note 691.0, 29-Mar-1995
Byrne: PC Motherboards
----------------------
Someone here has a 386 that instead of tossing 
they want to get a 486 motherboard for.
(what an odd sentence that is).
    
There are numerous vendors in the Computer Shopper.
    
Any comments on this subject welcome such as good/bad vendors,
gotchas, etc.



Note 691.1, 29-Mar-1995
Kennedy: 
---------
  What things do they want to save from their old system? The old system
probably doesn't have an advanced bus, so you'd want a new video card and
hard disk controller. The new motherboard may not use the same memory as
the old system, etc.

  I'd carefully price the swap after including all the other parts that
were needed and compare it against a new system (without keyboard and mon-
itor if that's the way you want it) and see how close it comes. 



Note 691.2, 4-Apr-1995
Hanrahan: 
----------
True, Terry, but then on the other hand, a lot of those low-priced 
complete systems have really poor motherboards, video cards, etc., in them.  

At least if you build a system up from bits, you know what you're getting.



Note 691.3, 4-Apr-1995
Kennedy: 
---------
  I was thinking of "name brand" systems, but you're right - even then you
can get some real junk.

> At least if you build a system up from bits, you know what you're getting.

  Well, you know what brands/models you're getting. Unless you've got exper-
ience with those exact items, you may still be in for a big surprise.

































The DECUServe Journal  July, 1995                              Page 15
PC Motherboard Replacement


Note 691.4, 6-Jun-1995
Killeen: 
---------
Assuming one gets a good upgrade motherboard are there any issues with 
mounting it i.e. are the mounting holes standard?



Note 691.5, 6-Jun-1995
Hanrahan: 
----------
Don't get me started... they are close to standard.  However I have on
several occasions encountered "standard" MBs that won't *quite* fit on
top of the mounting holes in "standard" cases.  Or else the MB fits but
then you find that the hold-down screws for the cards in the slots need
to be extended a bit.  And so on...



Note 691.6, 6-Jun-1995
Hahn: No structural porblem
---------------------------
Of the 20 or so various board upgrade I have put together I have 
not had problems on remounting the new board.  A couple of boards
did not match one or two holes
        


Note 691.7, 6-Jun-1995
Kennedy: 
---------
> Assuming one gets a good upgrade motherboard are there any issues with 
> mounting it i.e. are the mounting holes standard?

  "It depends". If you have a regular-footprint system it has chance of
working. If you currently have a slimline system (like a Northgate) which
has video, serial, parallel, etc. on the motherboard it's unlikely to fit.

  OTOH, a new case/power supply isn't that expensive (a top-of-the-line
case, PS, and overtemperature sensor from PC Power & Cooling is about $350,
and the typical "economy" case+PS is < $100).

  More important is picking a good motherboard. Many of them are marginal
designs and will fail under non-DOS+Windows OS's like Unix, Netware servers,
Windows NT, and Windows 95.

  What processor type/speed and bus are you looking at?





































The DECUServe Journal  July, 1995                              Page 16
PC Motherboard Replacement


Note 691.8, 6-Jun-1995
Killeen: WNT and WIN95 system
-----------------------------
> What processor type/speed and bus are you looking at?

P90/100 and PCI.

The current system is a 486 tower - the motherboard has an IBM name.



Note 691.9, 6-Jun-1995
Kennedy: 
---------
  In that case I would recommend the Intel Premier PCI II motherboard (also
known as the "Plato" board. It's a P5-90 w/ 3 PCI slots and 5 ISA slots. I
use these with the BusLogic 946C SCSI controller (the MB has 2 IDE controllers
on board, along w/ serial (2), parallel, floppy).

  Be careful with memory on these - you need tin-lead 72-pin parts with parity
(N x 36) that support fast page mode. You can get some real junk SIMMs if you
aren't careful.

  I have a source for the boards and SIMMs if you're interested.

  Stay away from Pentium motherboards that have VLB slots, especially in com-
bination with PCI. Also, beware of Intel's newer motherboards based on the
Triton chipset - it doesn't support parity.




  Testing for Removable Disk
  ------- --- --------- ----

  Abstract:

  When dealing with removable disks, it's sometimes useful to be able
  to determine from a program (i.e., not by walking up to the drive
  and peering inside) whether or not a volume is present.  It also be
  nice at times to do this in less time than it takes MOUNT to fail
  when nothing's in the drive.  An alternative is discussed below.



  Participants:

  John Briggs, Scott Harrod.

  Conference:  VMS


































The DECUServe Journal  July, 1995                              Page 17
Testing for Removable Disk


Note 2536.0, 8-Jun-1995
Harrod: Is it possible to determine if a removeable disk is present
-------------------------------------------------------------------
Is there a way to check, either from DCL or within a program, for the
presence of a removeable disk? We have a SCSI optical disk and would like
to determine if a cartridge is in the drive. 

The present kludge is to try to mount the disk. If there is no disk, the
mount will eventually fail; however, the timeout period has gone from
about a minute under VAX VMS 5.5-2 to about 4 minutes under Alpha VMS
6.1. Is there a way to specify that timeout? 



Note 2536.1, 8-Jun-1995
Briggs: PACKACK to determine if media is there
----------------------------------------------
Try an IO$_PACKACK QIO.  Completes instantly.  Either with an error code
or success indication.  It's a physical I/O and therefore requires
PHYS_IO privilege.

Warning:  DEC has an undocumented requirement that you do an IO$_AVAILABLE
          $QIO after a successful PACKACK if you want to put the volume
          back the way it was.  The PACKACK sets the volume valid bit.
          The IO$_AVAILABLE clears it.

          Ordinarily, you can get away without doing the IO$_AVAILABLE.
          However, certain tape devices (TUDRIVER) in combination with
          certain software (TAPESYS and SLS) can exhibit problems associated
          with this issue.

          Yes, I know you're worried about disk.  The same QIO works.

Example:

chan:
        .word   0
device:
        .ascid  /device/
iosb:
        .blkq   1
buff:
        .blkb   200
        .entry  check_disk,^m<>
        $assign_s chan=chan,devnam=device
        blbc    r0,100$
        $qiow_s chan=chan,func=#io$_packack,iosb=iosb
        blbc    r0,100$
        movzwl  iosb,r0
100$:
        $exit_s r0
        .end    check_disk

Worth every penny you paid for it.






























The DECUServe Journal  July, 1995                              Page 18
Testing for Removable Disk





Note 2536.2, 15-Jun-1995
Harrod: MSCP served vs. Locally attached
----------------------------------------
Thanks for the pointer to PACKACK. I tried it, but while the QIOW does
returns faster then a mount command if there is no disk in the drive,
it still takes some time. In fact in our testing we found that the
times were different depending on whether the disk was locally attached
or MSCP served from another node in the cluster. 

If the optical drive is attached to NODE2 and MSCP served to the cluster and
does *not* contain a cartridge, the time for the IO$_PACKACK QIOW function
call to complete and for the MOUNT command to timeout are as follows:
   

                                          NODE1          NODE2

       PACKACK QIOW                       2 min.         30 sec.
       MOUNT/NOASSIST/OVER=ID             4 min.          1 min.

Why does it take four times as long on the remote node. (I see by
the I/O manual that the MOUNT and INITIALIZE commands both do a 
IO$_PACKACK function call. It looks like MOUNT tries again if it fails
the first time.)

If there is a disk in the drive, the QIO and MOUNT command complete
rightaway.



Note 2536.3, 15-Jun-1995
Briggs: Try an async PACKACK
----------------------------
I've got no idea why it would take so long on the remote node.  Those
times seem outrageous to me.  But the only place I've ever played with
PACKACK is on tapes, locally and INFOserver attached.  In that environment,
the QIO completes very rapidly -- one second or less regardless of whether
the reel of tape is present.

Hmmm.  It seems that you do have a rapid way of determining whether
the disk is mounted.  Use an asynchronous QIO.  If it completes rapidly
and with success, the media is there.  If it fails to complete rapidly
or completes with failure, the media is absent.

All you have to do is set a timer and $CANCEL your PACKACK if the timer
expires.  You could also decide not to bother with the $CANCEL and let
image rundown handle it for you.

I haven't actually tried this.

































The DECUServe Journal  July, 1995                              Page 19
Testing for Removable Disk


Note 2536.4, 16-Jun-1995
Harrod: Thanks for the suggestion
---------------------------------
> Hmmm.  It seems that you do have a rapid way of determining whether
> the disk is mounted.  Use an asynchronous QIO.  If it completes rapidly
> and with success, the media is there.  If it fails to complete rapidly
> or completes with failure, the media is absent.

Yes, that sounds like it ought to work. I may give it a try. 
Thanks for the suggestion.




  Force Logging of a Session
  ----- ------- -- - -------

  Abstract:

  The problem at hand is how to ensure that a certain account's
  activity is always logged.  Of course, in the case of a
  highly-privileged account (as here), it would be difficult or
  impossible to overcome active malice, but there are other reasons
  why one might want to do such a thing, including the time-honored
  princliple of CYA.



  Participants:

  Dale Coy, Linwood Ferguson, Pierre Hahn, Mark Lasoff, Laurie
  Maytrott.

  Conference:  VMS


Note 2541.0, 20-Jun-1995
Maytrott: How can I force logging of an interactive login?
----------------------------------------------------------
I would like to create an account for which all activity is logged to a
file.  This is a priv'd operator account to do some of the higher level
stuff that's not done on a regular basis.  The "other account" aspect
will be used to create a "higher awareness" that he's working in a
more priv'd environment, and will add the logging to cover both of us
(the more "assertive" users can't claim he did something he didn't, and
it makes it easier to fix if a boo-boo does occur since we'll know
what was done).
    
I figured that SET HOST/LOG would pretty much do what I want.  I know I've
forgotten the /LOG a couple of times when I've meant to use it for a 
software install/upgrade, so I'd rather make it a more formal procedure 
than just having him do the SET HOST himself.  I've thought about creating a
symbol which would run a command file asking which node he wants, then
SET HOST/LOG to the desired node.






























The DECUServe Journal  July, 1995                              Page 20
Force Logging of a Session


    
My only hangup with this is that I can't think of a way to do a check
to MAKE SURE that the priv'd account is running with /LOG.  There isn't
any way for me to check this in LOGIN.COM, is there?  Or is there some
other forced log I can set up in LOGIN.COM instead of the SET HOST/LOG
method?  (other than using a Peek & Spy type s/w package?)



Note 2541.1, 20-Jun-1995
Hahn: See if the .log file was created
--------------------------------------



Note 2541.2, 20-Jun-1995
Coy: It may be worth doing, but...
----------------------------------
Notes:
    
  1.  You can't guarantee that a privileged account didn't subvert
your procedure (e.g., change the log file, do something outside the
logging, etc.)
    
  2.  SET HOST/LOG may not be terribly useful for screen-oriented
utilities.
    
  3.  When you "modernize" to use client/server stuff, the technique
will be even less useful.
    


Note 2541.3, 20-Jun-1995
Lasoff: Try PHOTO
-----------------
Try the PHOTO utility on one of the VS CD's mounted here.  You could
use some code in the LOGIN.COM that looks like this;
    
    $ if f$mode() .eqs. "INTERACTIVE"
    $    then !Create a photo log session.                 
    $         logname = "sys$login:" + -
                        (f$cvtime() - "-" - "-" - " " - ":" - ":" - ".")
    $         photo /log='logname
    $         mail /noself 'logname MAYTROTT /subj="PHOTO logfile."
    $         logoutt  !Once use logs-out of photo, then logout this session.
    $ endif
    


Note 2541.4, 20-Jun-1995
Maytrott: Thanks, all!
----------------------
I realize the limitations.  I'll try to find that program on here
later.






























The DECUServe Journal  July, 1995                              Page 21
Force Logging of a Session


    


Note 2541.5, 20-Jun-1995
Ferguson: VERB -> CLD and make it required?
-------------------------------------------
I agree with Dale's comments.  As a general rule, a privileged account
can do anything -- just so you know that ...
    
If you want to keep them from _forgetting_ to use /LOG, you might try
getting VERB and the SET CLD out, off and changing it to make /LOG
required, produce a command table for that user only. Then  a SET HOST
without /LOG will complain.   
    
I've not tried it, but it seems possible.  But again, ANYTHING you do
on a privileged account other than making it completely captive and
closely controlled will allow someone to defeat whatever auditing you
put on it.



Note 2541.6, 20-Jun-1995
Lasoff: Also try WATCH6
-----------------------
I forgot to mention the WATCH utility.  Version 6 (WATCH6) has the
ability to run in batch.  This will allow you to log activity into the
DCL logfile from a submit/keep/log command.  You won't be able to watch 
any process that has the UCB bit set (RTA devices from SET HOST).
    
This should be on one of the VS CD's here too.
    
Also, you may want to execute nightly ACCOUNTING and ANALYZE/AUDIT
reports to see if anybody is messing with the system.  I would also do
a directory/date=all of sys$manager:accountng.dat, and
sys$manager:security_audit.audit$journal to see if new versions exist.
    


Note 2541.7, 24-Jun-1995
Maytrott: Went with "simple" approach for now
---------------------------------------------
Perhaps I implied a bit too much of a hint towards a security slant
when I wrote the original message.  I get the feeling some of you think
that I think I can "secure" a fully priv'd user simply by "making" him log
his session.  ;-)  No, I wasn't born yesterday.... I realize that a priv'd
user is a priv'd user.  I just know that I have forgotten to log
sessions I've intended to log several times and wanted to make it
harder for him to forget.  
    
The reasons for logging were twofold... one, to do a CYA for him. We
have users who tend to be blamers.  If anything they do doesn't work
right, they want us to "prove" it's not something WE did before they
try to figure out what THEY did wrong.  I'm not saying that's what we
always need to do, but that's what they'd like... sometimes, it is






























The DECUServe Journal  July, 1995                              Page 22
Force Logging of a Session


easier for us to do a bit of that.   Only the developers get away with
that at all, and only because the development/data/test cluster has been 
and will be undergoing tremendous modifications (h/w, s/w, n/w).  The
operator would like to be able to clear himself if someone "accuses"
him of messing up something in their account while he performed the
[your favorite user support function here] on their account. 
    
Secondly, if something does accidently get messed up, I'd like to be
able to review the steps that led to the mess-up, both to come up with
a fix in a minimum amount of time, and to take some preventative
measures for next time (which might include setting up a .COM file or a
symbol to help perfom the function, or doing some additional training).
                  
The operator's in agreement with this. I can pretty much backtrack his
workload to find out what stuff he was doing when something messed up
if I had to anyhow, but would rather have the exact keystrokes when
possible.  I'm not trying to head off any malicious damage, because
there's enough other ways he could do stuff if he wanted... jeeez, he's
got physical access to the computers.  Doesn't take much checking to
find out how to get all the privs you want via physical access.  ;-)
    
I ended up creating a symbol which calls a command file.  The command
file asks which node he wants, then does a SET HOST/LOG to that node,
inserting the priv'd account username, but not password.  Not fancy,
and admittedly, not "secure" in restricting only logged access. 
However, it sports a big reminder banner upon login in case he forgets
to go in via the symbol (unlikely, since he won't be used to logging in
directly and typing in that username... the symbol doesn't match the
username).   I think this will do the trick for most of what I'm trying
to accomplish.  Thanks for all the pointers, though.  If I change my
mind and want to go to something more "airtight," I'll know where to
refer.




  SYSUAF in Mixed VMSclusters
  ------ -- ----- -----------

  Abstract:

  Mixed-architecture VMSclusters with a common SYSUAF file can pose an
  interesting problem:  certain user quotas may be precisely the right
  thing on a tuned-til-it-squeaks VAX, but make no sense whatsoever on
  a newly decked-out Alpha in the same cluster.  This is the topic of
  the next series of notes.



  Participants:

  Bruce Bowler, Rob Brooks, Chuck Gaffron, Jamie Hanrahan, David
  Kindler, Lynda Peach, Benjamin Pena, Pat Scopelliti, Patrick Stair,
  Glenn Zorn.






























The DECUServe Journal  July, 1995                              Page 23
SYSUAF in Mixed VMSclusters


  Conference:  VMS


Note 2533.0, 25-May-1995
Kindler: Sysuaf in axp/vax cluster
----------------------------------
I have installed our Alpha 2100 running vms6.1 into our vms 5.5-2
cluster.  Everything looks good but I would like to use our old
sysuaf.dat and rightslist.dat instead of having to recreate all new
accounts with new uaf parameters and identifiers.  
Does anyone have any experience with this?
    


Note 2533.1, 25-May-1995
Zorn: Should be no problem...
-----------------------------
You can do this no problem. I am currently running a mixed cluster and
sharing all the database files. (Your only problem will be one of the
net proxy files I believe) You might also have to change some quotas,
increase for the AXP, software.



Note 2533.2, 25-May-1995
Gaffron: not been a problem
---------------------------
We have been running our 2100s (AXP Vms V6.1) in the cluster with
all of our Vax Vms V5.5-2 system and they all share the same uaf,
rightslist, cdd dictionary (all users have sub-dictionaries on their
own dasd) and vmsmail_profile for the last 6 months. You should also
share the same lmf$license database or you will run into "license
units" problems like you would not imagine. (you should manage the
license database from the Vms V6.1 nodes)
    


Note 2533.3, 25-May-1995
Brooks: 
--------
Just for completeness, the proper way to share SYSUAF and RIGHTSLIST
(or more accurately, the way to move SYSUAF and RIGHTSLIST from
their default location of SYS$SPECIFIC:[SYSEXE]) is to define
executive mode logicals that point to their new locations.
    
The logicals are named (logically enough) SYSUAF and RIGHTSLIST.

You probably knew this anyway, but just in case . . .




































The DECUServe Journal  July, 1995                              Page 24
SYSUAF in Mixed VMSclusters


Note 2533.4, 25-May-1995
Kindler: More confusion - sysuaf for vax/axp
--------------------------------------------
I guess I am a little confused.  I was told by someone from DEC that I
should have seperate sysuaf files because the uaf parameters need to be
higher for the alpha.  I am reading the guidelines for vmscluster
configurations (vms 6.1) and it says (on page 8-11) that you SHOULD use
the same sysuaf file but that you need to change some system
parameters.  It does not tell you what system parameters to change. 
Can anyone point me in the right direction??
    


Note 2533.5, 25-May-1995
Scopelliti: You say pages, I say pagelets
-----------------------------------------
My guess is they meant the PQL parameters.  Be aware of the distinction
in the use of the term "page" between VAX and Alpha architectures.  A
VAX page is 512 bytes, while an Alpha page is (currently) 8192 bytes,
or 16 "pagelets".  UAF and PQL entries on both systems deal in pagelets.
    
Now if the sysuaf.dat file had nodename-specific settings...



Note 2533.6, 27-May-1995
Stair: Use PQL_M* and certain logicals
--------------------------------------
We use a cluster-wide set of SYSUAF, net-stuff, and other files.  My 
reading of the documentation indicates that using a single SYSUAF file 
has the potential for some non-optimized parameters, depending on your 
usages and your settings of the proper PQL sysgen parameters.  The 
manual "A Comparison of System Management on OpoenVMS AXP and OpenVMS 
VAX" that came with AXP V1.5 (I haven't investigate the 6.1 edition if 
there is one), section 2.2.26 "How Process Quotas Are Determined in 
Dual-Architecture VMSclusters" describes how to use the PQL_M* on VAX 
and Alpha nodes in a mixed cluster to set minimum process quotas.  We 
have used that to overcome differences between the two architectures 
using the same UAF FILE.

What we do is define in SYLOGICALS.COM (on both the VAX and the Alpha 
bootups) various cluster-wide logicals.  Of particular interest here, 
we've define CLUSTER$BOOT$SYSEXE to point to the _VAX_ (not the Alpha, 
because the VAX boots first at our site) SYS$COMMON:[SYSEXE] directory.

Then, each SYSTARTUP_VMS.COM (both VAX and Alpha bootups) have the 
following (which, I suppose, could just as well be in SYLOGICALS.COM):

$ DEFINE/SYSTEM/EXEC SYSUAF          CLUSTER$BOOT$SYSEXE:SYSUAF
$ DEFINE/SYSTEM/EXEC RIGHTSLIST      CLUSTER$BOOT$SYSEXE:RIGHTSLIST
$ DEFINE/SYSTEM/EXEC NETPROXY        CLUSTER$BOOT$SYSEXE:NETPROXY
$ DEFINE/SYSTEM/EXEC NET$PROXY       CLUSTER$BOOT$SYSEXE:NET$PROXY
$ DEFINE/SYSTEM/EXEC NETOBJECT       CLUSTER$BOOT$SYSEXE:NETOBJECT
$ DEFINE/SYSTEM/EXEC VMSMAIL_PROFILE CLUSTER$BOOT$SYSEXE:VMSMAIL_PROFILE






























The DECUServe Journal  July, 1995                              Page 25
SYSUAF in Mixed VMSclusters


$ DEFINE/SYSTEM/EXEC SYS$SYLOGIN     CLUSTER$BOOT$SYSMGR:SYLOGIN.COM



Note 2533.7, 1-Jun-1995
Peach: UAF> HELP on WS params wrong for 6.1
-------------------------------------------
Also be aware that the HELP under UAF> (Authorize) has *incorrect*
information about the AXP settings.  This is VMS version 6.1. 
    


Note 2533.8, 6-Jun-1995
Pena: want a cheesy hack? ;-)
-----------------------------
One way we did it was to leave all process quotas at the VAX settings,
and in sylogin (or some such place, can't recall offhand at the
moment), run a home-brewed kernel mode hack that raises the appropriate
quotas as needed on the AXP. Ta-da! Problem solved. :)
    
'Course we also bumped up the PQL_M* parameters on the AXP.
    


Note 2533.9, 8-Jun-1995
Hanrahan: 
----------
> 'Course we also bumped up the PQL_M* parameters on the AXP.

That was what I was about to suggest.  Why did you need the program 
also?



Note 2533.10, 8-Jun-1995
Pena: 'twas the user process quotas that needed bumping up
----------------------------------------------------------
Ah. PQL_M* bumped on the AXP wasn't a problem. However, in a mixed
architecture cluster (1 AXP, 7 VAX, all running 6.1), we weren't sure
how to differentiate user process quotas under VAX or AXP with a common
UAF for all nodes in the cluster, so we went by the route of the hack
to bump up user quotas upon login to the AXP.



Note 2533.11, 8-Jun-1995
Bowler: but...
--------------
But the processes will get bumped up to PQL_M* automatically (except
under certain conditions, which only occur during startup).  Unless
the hack bumps different users differently, it's not needed.

































The DECUServe Journal  July, 1995                              Page 26
SYSUAF in Mixed VMSclusters


Note 2533.12, 8-Jun-1995
Pena: memory's bad today. :)
----------------------------
You're right, sorry.
    
The AXP joined the cluster about a year ago... owas being set up for
about two months instead of the usual 1, because of the changes in
paradigm we had to use, as far as AXP planning goes
(hardware/software/migration). During this time, system support
determined the way to go about this addition was:
    
Bump up all user quotas in the clusterwide UAF to AXP levels, then in
SYSGEN on the VAXes, bump necessary quotas down, as SYSGEN quotas
having precedence over UAF quotas. Initially, we determined a kernel
mode hack was necessary to handle certain quotas on the VAXes (can't
recall what at the moment), although I don't know if we are still using
the hack or not. I can check around to see if we came up with a cleaner
solution than the hack.
    



  MOUNT and OpenVMS V6.1
  ----- --- ------- ----

  Abstract:

  "Hmm, I wonder what this cryptic bit in the release notes means?"

  "What do you mean, I can't mount my own disk?"

  If you find the above (hypothetical) monologue at all disturbing,
  you may want to read the next few notes.  They discuss changes to
  MOUNT in OpenVMS V6.1, and the new SET SECURITY command.



  Participants:

  Larry Clegg, Larry Kilgallen, Charlie Wrenn.

  Conference:  VMS


Note 2544.0, 22-Jun-1995
Wrenn: MOUNT under OpenVMS/VAX 6.1
----------------------------------
The OpenVMS 6.1 new features manual indicates that the MOUNT
command has been changed, but neither that nor the release
notes tell what it is that is now different (that I can
find  anyway). Does anyone know how MOUNT has been changed,
particularly with respect to MOUNT/CLUSTER?
































The DECUServe Journal  July, 1995                              Page 27
MOUNT and OpenVMS V6.1


Note 2544.1, 22-Jun-1995
Wrenn: Can't mount my own disk
------------------------------
Following up my own question with more detail:

When I first asked, it was mostly out of curiosity, but now I
really do have a problem. This was working before the OpenVMS/VAX
6.1 upgrade (from 5.5-2).

I have an account that occasionally needs to mount a removable,
optical disk on the cluster. By giving this account SYSNAM priv,
it had worked fine. Now when I try to mount the disk (whether
cluster or not), I get:

%MOUNT-F-NOPRIV, insufficient privilege or object protection violation

Now that appears to be a prefectly lucid error message, but I'm just
not understanding. The disk volume is owned by this account. What
privilege does it need? So far my only work around has been to 
do the mount from the system account, which is not an acceptable solution.
I'm confused.



Note 2544.2, 22-Jun-1995
Clegg: USE SET SECURITY
-----------------------
I encountered this same problem.  It doesn't matter who owns the volume
because the real problem is accessing the device itself.  Digital has
increased the security awarness in VMS V6.1.  I had to go through all my tape
devices which the "public" can mount/dismount and set the correct protection
levels.  For my environment I used: 

  $ SET SECURITY/DEVICE/PROT=W:RWLP $1$MUB0:

I then found out that this enables the device only temporarily.  You'll lose 
this setting next time you reboot unless you do:

  $ SET SECURITY DEVICE/CLASS=SECURITY_CLASS -
  _$ /PROFILE=TEMPLATE=TAPE /PROT=(W...)

Be warned:  this will set it for ALL device class = tape units.  This was ok 
in my shop but it might not be ok in your shop.  The HELP SET SECURITY is not 
too clear about all this.  A DSNlink Q & A got me going in the right 
direction.
 


Note 2544.3, 22-Jun-1995
Clegg: SET SECURITY is new
--------------------------
I forgot to note that SET SECURITY is new to 6.1 and supercedes SET PROT.
Enjoy.































The DECUServe Journal  July, 1995                              Page 28
MOUNT and OpenVMS V6.1


Note 2544.4, 22-Jun-1995
Clegg: More TMSCP MOUNT Problems
--------------------------------
Well I can't seem to get all my thoughts into one place.

I also have had LOTS of problems with MOUNTING a TMSCP-served tape drive under
OpenVMS/VAX 6.1.  If I attempt to MOUNT a tape drive which is TMSCP served to
the cluster I will hang in the MOUNT command.  The only way of freeing
everything up is to reboot the whole cluster!  That's right - the whole
cluster.  Just rebooting the node to which the tape drive, a TU81+ in this
case, is physically attached does no good.  Rebooting the node which attempted
to mount the tape drive does no good either.  The problem only clears up with
a full cluster reboot.  Digital thinks that I may have had a unique
hardware/software problem.  My TU81+ has been effectively gutted and replaced
with new parts.  Digital has also downloaded a brand new TMSCP process.  They
say I'm the only one with this version of TMSCP - what an honor.  I haven't
had a chance to reboot the whole cluster yet to implement this new TMSCP -
that will happen this weekend.  As long as I access the tape drive only from
it's host node everything is a-ok. 

Digital tells me this problem has surfaced at only one other customer site but 
they can't reproduce as frequently as we can.



Note 2544.5, 23-Jun-1995
Wrenn: SET SECURITY does it
---------------------------
> $ SET SECURITY/DEVICE/PROT=W:RWLP $1$MUB0:

Thanks, Larry, That did it.  Actually it's 
/CLASS=DEVICE, but I figured that out. 



Note 2544.6, 23-Jun-1995
Kilgallen: SYS$EXAMPLES:RESET_DEVICE_PROTECTION.COM
---------------------------------------------------
> I then found out that this enables the device only temporarily.  You'll lose 
> this setting next time you reboot unless you do:
> 
>       $ SET SECURITY DEVICE/CLASS=SECURITY_CLASS -
>        _$ /PROFILE=TEMPLATE=TAPE /PROT=(W...)

Note that DEC has provided a command procedure to do this:

Directory SYS$COMMON:[SYSHLP.EXAMPLES]

RESET_DEVICE_PROTECTION.COM;1               6

Total of 1 file, 6 blocks.

































The DECUServe Journal  July, 1995                              Page 29
Windows NT Crash Dumps


  Windows NT Crash Dumps
  ------- -- ----- -----

  Abstract:

  By now, anyone who hasn't been hiding under a very large and sturdy
  rock has heard that OpenVMS and Windows NT have many commonalities
  of design, and indeed of designer.  [Editorial Digression:  If we
  hear the smug little ditty about VMS incremented giving WNT one more
  time, we just might do something drastic.] But crash dump analysis?
  Why not?  It might not be quite at the level of the System Dump
  Analyzer yet, but it can be done.



  Participants:

  Jamie Hanrahan, Ken Johnson.

  Conference:  WINDOWS_NT


Note 95.0, 27-Apr-1995
Johnson: Analyzing Crash Dumps
------------------------------
Using WinDbg KD to examine crash dumps for the standard Intel build of
Windows NT 3.5, using an Intel system running the same build.

There may be shortcuts or simplifications that could be made to some of
these steps.  However, when I do it this way, it works for me.  I tried 
lots of other ways that didn't work.

1)  Copy the \support\debug\i386... subtree from the distribution CD to
a convenient working directory and subtrees on a hard drive, say
YOURDISK:\support\debug\i386. 

2)  Copy \mstools\bin\i386\*.dll from the WNT 3.5 SDK to
YOURDISK:\support\debug\i386. 

3)  Copy \mstools\help\windbg.hlp from the WNT 3.5 SDK to
YOURDISK:\support\debug\i386. 

4)  Open a command prompt window and set the default directory to
YOURDISK:\support\debug\i386. 

5)  Type WINDBG at the command prompt. 

6)  Using the 'User Dlls' choice from the 'Options' menu of WinDbg, set
the Symbol Search Path to YOURDISK:\support\debug\i386\symbols. 

7)  Using the 'Kernel Debugger' choice from the 'Options'menu of WinDbg,
open the crash dump file. 

8)  Give the 'L' command in the command window of the debugger -- this






























The DECUServe Journal  July, 1995                              Page 30
Windows NT Crash Dumps


gets things loaded. 

9)  Give the '!drivers' command in the command window of the debugger --
this prints the starting address of the loaded drivers. 

10)  Give the 'K' command in the command window of the debugger -- gives
a very limited traceback of the stack. 

11)  Give the 'DD @ESP' command in the command window, followed by 'DD'
commands as needed. This dumps the stack as longwords. 



Note 95.1, 29-Apr-1995
Hanrahan: Some comments...
--------------------------
Pretty good job!  Just a few pointers:

> 1)  Copy the \support\debug\i386... subtree from the distribution CD to
> a convenient working directory and subtrees on a hard drive, say
> YOURDISK:\support\debug\i386. 

You don't really need all those symbol files.  I usually get by with just the 
drivers, ntoskrnl, and hal. 

> 2)  Copy \mstools\bin\i386\*.dll from the WNT 3.5 SDK to
> YOURDISK:\support\debug\i386. 
> 
> 3)  Copy \mstools\help\windbg.hlp from the WNT 3.5 SDK to
> YOURDISK:\support\debug\i386. 

If you've installed the SDK to YOURDISK:, you already have done effectively
this -- just click on the WINDBG icon in the SDK Tools group. 

EXCEPT.... you have to be using the kdext*.dll that is built for the 
"build" you're debugging.  If you installed NT from the SDk CD-ROM, 
fine.  Otherwise you need to go to the CD-ROM from which you installed
NT and get the kdext*.dll from there.  If you are debugging multiple
platforms, ie free and checked build, you have a problem -- they're 
named the same and there is no way in windbg to point it at a particular 
kdext dll.  But, WINDBG looks in the directory in which its own .exe
lives to find its dlls.  So you can create a couple of different
directories, say one for free build and one for checked, and put the
appropriate windbg.exe and kdext*.dll's in each.   

> 4)  Open a command prompt window and set the default directory to
> YOURDISK:\support\debug\i386. 
> 
> 5)  Type WINDBG at the command prompt. 

Or just go to file manager, open the directory, and click on windbg.exe.

> 6)  Using the 'User Dlls' choice from the 'Options' menu of WinDbg, set
> the Symbol Search Path to YOURDISK:\support\debug\i386\symbols. 






























The DECUServe Journal  July, 1995                              Page 31
Windows NT Crash Dumps



similar comments for the kdext*.dll's apply to the symbol files also.
ie you may want to have ...i386\free807\symbols and ...i386\chkd807\symbols .

Note that the upper-level directory names for all of these things are
not critical.  Mine are in \kdbg\free807\i386\symbols... and 
\kdbg\free807\i306\bin . 

> 7)  Using the 'Kernel Debugger' choice from the 'Options'menu of WinDbg,
> open the crash dump file. 
> 
> 8)  Give the 'L' command in the command window of the debugger -- this
> gets things loaded. 
> 
> 9)  Give the '!drivers' command in the command window of the debugger --
> this prints the starting address of the loaded drivers. 
> 
> 10)  Give the 'K' command in the command window of the debugger -- gives
> a very limited traceback of the stack. 

You can also open the "calls" and "locals" windows under the "windows"
menu. 

!help at the prompt in the command window gets a list of the kernel debugger
extension commands.  Lotsa neat stuff in there...

Just as in a VMS crash dump, the stuff at the very top of the stack will
usually be in the routine that raised the bugcheck exception.  You have to look
down the stack a bit to find out who really caused the problem. 



Note 95.2, 4-May-1995
Johnson: Cross-platform dumps; page file; Stop screen
-----------------------------------------------------
Note that in some cases, the crash dump is written to the page file at 
the time of the crash, and copied to the dumpfile when the system next 
boots.  So if your checked build system crashes, and you chose the free 
build when you reboot, the dump file might or might not be there.  You 
may need to reboot the build that crashed and let it come up completely 
so that the dump can get copied.  It might be that this is only a 
problem if the dump file is not in the default location.  On one system, 
I needed to enlarge the page file so it would come up without 
complaining about virtual memory shortages when coming up after a dump.

I have tried to examine MIPS and Alpha crashdumps from an Intel system.  
I couldn't open the MIPS dumpfile at all -- reports "Crash Dump 
Specified is not really a crash dump".  I could open the Alpha dumpfile, 
but some commands, like !drivers, wouldn't work.  For now, I recommend 
using a system of the same architecture to examine the dumpfile.

> Just as in a VMS crash dump, the stuff at the very top of the stack will
> usually be in the routine that raised the bugcheck exception.  You have
> to look down the stack a bit to find out who really caused the problem. 






























The DECUServe Journal  July, 1995                              Page 32
Windows NT Crash Dumps



Jamie's point here is a good one.  It moves the discussion from the 
mechanics of being able to look at a crash dump (the part that was hard 
for us to figure out, even with some calls to Microsoft support) to 
interpreting the crash dump.

A similar caution applies to interpreting the Stop screen display (Blue
Screen of Death).  At least on the Intel platform, we sometimes see
addresses displayed from the stack which are leftovers, and not in the
current call chain.  We usually see the actual address of the fault, but
sometimes don't see the return address from the current routine -- which
is a bother if the fault occurs in a special bug-trapping routine which
is checking for a corrupted database or inconsistent state.  (Which got 
us into looking at crash dumps in the first place.)



Note 95.3, 4-May-1995
Hanrahan: 
----------
> For now, I recommend 
> using a system of the same architecture to examine the dumpfile.

That's odd.  Cross-platform kernel debugging certainly works.  You have
to select the right processor architecture in one of the Options... dialogs.




  SET SECURITY on Tape Units
  --- -------- -- ---- -----

  Abstract:

  The new SET SECURITY command, also discussed elsewhere in this
  issue, can change the default protection for specific tape (and
  other) units.  The question in the following stream concerns whether
  or not those settings survive a reboot.



  Participants:

  Debbie Driesman, Alan Hunt, Larry Kilgallen, Laurie Maytrott,
  Harrison Spain.

  Conference:  SECURITY


Note 309.0, 5-Jun-1995
Maytrott: How permanent is SET SECURITY on tape units?
------------------------------------------------------
If I use SET SECURITY (VAX/VMS V6.1) to change the default protection
for a tape unit, does the new protection go into a permanent database






























The DECUServe Journal  July, 1995                              Page 33
SET SECURITY on Tape Units


and survive rebooting or does this have to be done at every boot-up?
    


Note 309.1, 5-Jun-1995
Spain: I think this needs to be done in the SYSTARTUP_VMS.COM file
------------------------------------------------------------------
I don't think the protections are retained since these devices are
created during reboot.  I just finished setting an ACL on the NET0:
device and it was no longer there after the reboot.
    
Adding the command to your SYSTARTUP_VMS.COM should do the trick :-).
    


Note 309.2, 5-Jun-1995
Kilgallen: 
-----------
For clusterwide devices, it is certainly supposed to be retained.

I thought it was for tapes as well.

Of course an individual driver can change the protection of its device
after UCB creation, and perhaps that is what is happening for NET0:.



Note 309.3, 6-Jun-1995
Driesman: Disk and tape security is permanent
---------------------------------------------
Section 4.3 of the V6.0 OpenVMS VAX Guide to System Security describes
when and where device security profiles are derived.  For disks and
tapes it states
"Once the device is visible within a cluster, its profile, with any
modifications, is retained across system restarts."  However, template
changes don't propagate down to devices, so the device security profile
needs to be re-set if the profile changes.



Note 309.4, 6-Jun-1995
Hunt: It can be permanent
-------------------------
From experience they can be made permanent.  Same for floppy drives.




  UNIX Filename Globbing
  ---- -------- --------


































The DECUServe Journal  July, 1995                              Page 34
UNIX Filename Globbing


  Abstract:

  Although it beats the heck out of your faithful editors what is
  meant by "filename globbing", we trust this information will be of
  use to someone.  (But then, our own operating system biases are well
  established.)

  The pointer to C-Kermit, incidentally, comes naturally to Terry
  Kennedy -- after all, he did the VMS implementation.



  Participants:

  Terry Kennedy, Howard Siegel, Ray Whitmer.

  Conference:  UNIX_OS


Note 276.0, 18-Jun-1995
Siegel: Filename globbing in a program
--------------------------------------
How does one do file name globbing from within a program?



Note 276.1, 22-Jun-1995
Whitmer: What I have seen was pretty messy on UNIX.
---------------------------------------------------
I assume that for some reason the automatic globbing from the shell is
inadequate.  I am no UNIX expert, but the code I have seen written by
others in UNIX is very messy, i.e. open the directory file, use
standard header definitions to know how to read that file, and pass
them through a wildcard-matching algorithm.



Note 276.2, 23-Jun-1995
Siegel: Not quite
-----------------
> I assume that for some reason the automatic globbing from the shell is
> inadequate.  

Not inadequate, just not available.  The program is not getting the globbed
filenames from the command line because that is not where the file names
are comming from, thus I must glob from within the program.  If I could
get the names from the command line then all my problems would be fixed.
I tried constructing a new command line and then 'execl'ing it, but that
doesn't go through the shell, so no filename globbing gets done. Oh well,
looks like I'm gonna have to get dirty doing the parsing myself.


































The DECUServe Journal  July, 1995                              Page 35
UNIX Filename Globbing


Note 276.3, 23-Jun-1995
Kennedy: C-Kermit
-----------------
  I'd suggest you take a look at C-Kermit's ckufio.c module. It can expand
a wildcarded list to individual names itself or call the shell to do it.
You may already have C-Kermit source, or you can get it from watsun.cc.col-
umbia.edu in kermit/c/.



Note 276.4, 24-Jun-1995
Siegel: found it
----------------
Thanks Terry.  Note that the path was actually kermit/c-kermit/ckufio.c.




  POP Server for Unix
  --- ------ --- ----

  Abstract:

  The following series of notes considers various ways and means of
  getting a POP (Post Office Protocol) server for remote mail clients
  to run under Unix.  Also included at the end is a worthwhile lession
  in security issues related to running as root.



  Participants:

  Kevin Angley, Charlie Byrne, Joe Matuscak.

  Conference:  UNIX_OS


Note 275.0, 13-Jun-1995
Byrne: POP Server for DEC Unix
------------------------------
What is a good POP server to use on DEC OSF V3.0?
    
How about the Popper from ftp.cc.berkeley.edu 
- that seems to be fairly popular (npi).
    
Does one come with the distribution CD's
(freeware or complementary products)?





































The DECUServe Journal  July, 1995                              Page 36
POP Server for Unix


Note 275.1, 20-Jun-1995
Matuscak: 
----------
I dont know about OSF, but Ive got an Ultrix system and I use the pop
server from ftp.qualcomm.com, the home of Eudora (A quite nice pop
client for Windows and Mac, BTW). My impression is that someone at
Qualcomm has been maintaining it.



Note 275.2, 21-Jun-1995
Byrne: mh includes pop
----------------------
Now I see that mh, the Rand Mail Hander, comes with the OSF/1 3.0 kit
(as an optional subset). This supposedly includes POP server
capabilities.

So now the question becomes, 
    
Is there any reason for me to use a different POP server 
(e.g. from ftp.qualcomm.com) if I already get one with mh?



Note 275.3, 21-Jun-1995
Matuscak: 
----------
My impression is that mh isnt all that mainstream, but other than that,
I dont have much idea. Is mh supported on OSF? (With Ultrix, its on the
unsupported kit).  BTW, if you want a pretty decent terminal oriented
mail program, check out pine. Its actually pretty nice.



Note 275.4, 23-Jun-1995
Byrne: PINE SOLution, get PINE TAR file (npi)
---------------------------------------------
Well actually it turns out that the Pine package 
(FTP.CAC.WASHINGTON.EDU) is exactly where I ended up getting my POP
server from. They also have IMAP server, the guy who wrote the
original IMAP is working there. Installation was simple, I was able to
copy the OSF/1 binary from another machine and then you just:
    
    1) Stick the binary (ipop3d) somewhere (eg /usr/local/bin)
    2) tell /etc/services that port 110 => pop3
    3) tell /etc/inetd.conf pop3 => /usr/local/bin/ipop3d

Dat all! Go back to your PC, fire up Eudora and you're all set.
    
I am Unix newbie and got stuck because:
    
1) when I FTP'ed the binary it apparently didn't copy the 
"Executable" permissions, so I had to CHMOD them (once I figured this
out)






























The DECUServe Journal  July, 1995                              Page 37
POP Server for Unix


    
2) If you are root and have the # prompt and you have program x in your
current directory, and you type "x", it sez "Not found". This drove me
nuts until I finally decided what the heck let me try this from a
regular user account, and haha it worked. I guess it has something to
do with root shell or something? That will teach me to stay in root
longer than necessary, eh?



Note 275.5, 23-Jun-1995
Angley: No . in root's path
---------------------------
The difference probably was that the regular user account has "." (the
current working directory) in the path, and the root account does not.
Having . in the path can be dangerous in that an unscrupulous user
could have a phony "ls" or some such in their directory and seek your
help as root. Thus, the ls you thought was ls is really their version
of ls (which, if they are clever, also looks like ls). The root account
may not have "." because COPS and other security checkups will
check and correct this in the root account.
    



  4mm DAT Autoloader
  --- --- ----------

  Abstract:

  A long-running stream on DEC 4mm DAT tape drives contains the
  following discussion of the SCSI arcana necessary to control the
  autoloader on a TLZ06L.



  Participants:

  David Campen, Jamie Hanrahan, Bob Hassinger, Terry Kennedy, Frank
  Nagy, Keith Parris.

  Conference:  HARDWARE_HELP


Note 1482.19, 31-May-1995
Hassinger: "SCSI Random Access Commands" (re: TLZ06L programming)
-----------------------------------------------------------------
Well, I finally have one of these TLZ06L DAT drives with autoloader in
hand (until it is swapped for the TLZ07L I ordered and which is on very
long lead time at the moment).  Seems pretty nice - at least compared
to the TU81+...

Anyway, I am now down to really needing information on controlling the
autoloader.  The only booklet that came with it says a grand total of






























The DECUServe Journal  July, 1995                              Page 38
4mm DAT Autoloader


the following: "In Random Access Mode, the loader responds to all of
the SCSI Random Access Commands".

I have never done any SCSI programming and have no idea where to find
what I need.  I tried the CSC, but so far they have no idea either.

A pointer to info on "the SCSI Random Access Commands", and some
examples of using them, particularly from VMS would be very helpful. 
    
Also, at this point I am unclear about the LUN issue Terry mentioned
back in .6.  It would seem there must be a way to address at least
twelve tapes since DEC sells software to use with this device and I am
under the impression it can use the larger magazines (standard is 4
tapes, there was an 8 tape magazine, and the 12 tape magazine is now
the standard large size).  Can anyone educate me a little on this? 



Note 1482.20, 31-May-1995
Kennedy: 
---------
  It's in the Archive (Connor) manual. If you call them they should be able
to get you the info. If you get desperate, you can always order the SCSI 
spec from Global Engineering Documents. I have a copy of the manual for the
Archive 4592NP (the older version) around somewhere and I could mail you a
copy of the loader section if you want.

> Also, at this point I am unclear about the LUN issue Terry mentioned
> back in .6.  It would seem there must be a way to address at least
> twelve tapes since DEC sells software to use with this device and I am
> under the impression it can use the larger magazines (standard is 4
> tapes, there was an 8 tape magazine, and the 12 tape magazine is now
> the standard large size).  Can anyone educate me a little on this? 

  SCSI "things" are disks, tapes, CD-ROMs, media loaders, etc. In theory,
a given device can't be more than one type. In order to support environments
with a strict interpretation of the SCSI spec, these drives claim to be
multiple devices - a tape drive, a media loader, and so forth. All of the
commands work on all of the devices, though. Each "device" is a LUN. On a
VAXstation with built-in SCSI, it'd show up as different devices (MKA400,
MKA410, etc., I guess). I don't know if DEC's driver actually knows what to
do with a media loader, though - it may not even configure it.

  Here's a simple backup script that runs a media loader using only normal
VMS commands. It works because the loader automatically inserts the next
tape (next slot) if you eject (dismount/unload) a mounted tape.

$ !'f$verify(0)
$ ! BACKUP.COM - Do the system backups
$ ! T1.0-00 - 03-Jan-90 - tmk - VMS V5.2, funky parameters
$ ! T1.1-02 - 10-Sep-90 - tmk - VMS V5.3-1, ask about each tape
$ ! T1.2-03 - 07-Apr-91 - tmk - VMS V5.4-1, add new devices
$ ! T1.2-04 - 21-Nov-91 - tmk - VMS V5.4-3, add new devices
$ ! T1.2-05 - 22-May-92 - tmk - Start of the Great Disk Reorganization






























The DECUServe Journal  July, 1995                              Page 39
4mm DAT Autoloader


$ ! T1.2-06 - 30-Nov-92 - tmk - 2 more RA90's, 2 HSC's
$ ! T1.2-07 - 04-Aug-93 - tmk - 2 more RA90's (now a full 8-pack 8-)
$ ! T1.2-08 - 14-Aug-94 - tmk - GDR II (DSSI's)
$ set noon
$ !
$ write sys$output ""
$ write sys$output "Working on $1$DIA0: on 4mm tape drive..."
$ show time
$ set process/name="Backup $1$DIA0:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dia0: mub0:dia0.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ write sys$output "Working on $1$DIA1: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup $1$DIA1:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dia1: mub0:dia1.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ write sys$output "Working on $1$DIA2: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup $1$DIA2:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dia2: mub0:dia2.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ write sys$output "Working on $1$DIA3: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup $1$DIA3:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dia3: mub0:dia3.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ write sys$output "Working on $1$DIA4: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup $1$DIA4:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dia4: mub0:dia4.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !






























The DECUServe Journal  July, 1995                              Page 40
4mm DAT Autoloader


$ write sys$output ""
$ write sys$output "Working on $1$DUA1: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup $1$DUA1:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  $1$dua1: mub0:dua1.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ write sys$output "Working on SPCVXZ$DUA0: on 4mm tape drive..."
$ wait 00:01
$ show time
$ set process/name="Backup SPCVXZ:"
$ backup/image/fast/crc/norecord/block=32256/ignore=interlock -
  spcvxz$dua0: mub0:dua0.bck/init/density=6250/rewind/noassist
$ show time
$ dismount mub0:
$ !
$ write sys$output ""
$ exit



Note 1482.21, 1-Jun-1995
Hassinger: Maybe I am the only one who wants to know...?
--------------------------------------------------------
Thanks Terry.  Yes, I think a procedure like that would work with the
TLZ06L since it has a similar automatic step to the next tape feature.

My main interest at this point is in the random access part because I
want to see if I can set it up to do something like use tape 1 on
Mondays, tape 2 on Tuesdays, etc.  That is, as near to an unattended
backup process as possible.  I currently do something along these lines
using a spare disk (I have described the particular scheme elsewhere on
DECUServe in the past).  As disk sizes keep going up and the amount of
day to day volatility increases with large data sets from the labs I am
looking for a new solution.

I am still unclear where DEC intendeds for me to look to find out
things like the LUN for the loader, as well as the commands.  There
seems to be a basic piece missing somewhere.  Maybe I have missed a
pointer to a manual?  I can't even find someone to ask.  It almost
seems as if DEC is trying to keep it a secret to force people to use
their proprietary backup software products.



Note 1482.22, 2-Jun-1995
Parris: 
--------
> Anyway, I am now down to really needing information on controlling the
> autoloader.  The only booklet that came with it says a grand total of






























The DECUServe Journal  July, 1995                              Page 41
4mm DAT Autoloader


> the following: "In Random Access Mode, the loader responds to all of
> the SCSI Random Access Commands".

This wording appears to simply be based on the SCSI-2 spec.  SCSI devices
come in various flavors, and different types can handle different sets of
SCSI commands.  A sample of the Peripheral Device Type table is:

        Code    Description
         00h    Direct-access device (e.g., magnetic disk)
         01h    Sequential-access device (e.g., magnetic tape)
         02h    Printer device
         03h    Processor device
         04h    Write-once device (e.g., some optical disks)
         05h    CD-ROM device
         06h    Scanner device
         07h    Optical memory device (e.g., some optical disks)
         08h    Medium changer device (e.g., jukeboxes)
         09h    Communications device

It appears your tape drive is a sequential device at heart, but also can handle
some additional random-access commands for the loader.

If one were starting from scratch, one would probably use GKDRIVER, the generic
SCSI Class Driver (or modify SYS$EXAMPLES:SKDRIVER.MAR) and send down the
appropriate SCSI commands to the tape drive to control the loader.  This sort
of SCSI work seems to be covered reasonably well in the VMS doc set. 

But I suspect this type of programming has already been done for that drive.
I'll check with some folks in my old group who work with SLS and PAB and try to
get an answer for you.



Note 1482.23, 2-Jun-1995
Hassinger: 
-----------
I have been getting some clues from here and there.  Right now it
appears that you need to do this through the MK driver since it already
has control of the tape drive (i.e. my drive is MKA500:).  As far as I
can tell now you talk to the driver with additional op codes and data
structures to controll the autoloader.  For example I was told
yesterday that op code A5h is "Move Medium"  which I think is supposed
to move tape cartridges back and forth between the slots in the
magazine and the drive by "address".  So far my information is too
sketchy to really understand or use it though.  I was only able to get
a little bit of the puzzle.

SLS sounds like a likely place to ask.  Everyone I ask keeps trying to
sell me the software products that use these drives.  They seem to be
the only people who have a clue how it works.


































The DECUServe Journal  July, 1995                              Page 42
4mm DAT Autoloader


Note 1482.24, 2-Jun-1995
Parris: 
--------
An answer from Rich Wrenn:

Yes, we don't do a very good job of helping users program the loaders.  
The answer to his question is that you use IO$_DIAGNOSE QIOs through the
GKDRIVER to LUN 1 of the device to pass standard SCSI Media Changer Interface
commands to the loader.  The standard SCSI commands are defined in the SCSI
standard which can be obtained from the American National Standards Institute.
I don't happen to have one on-line.  Documentation on the GKDRIVER can be found
in the VMS IO User's Guide.

We're working on a utility to help make this easy.  So, stay tuned.

Rich



Note 1482.25, 5-Jun-1995
Campen: This is why I haven't bought an autoloader.
---------------------------------------------------
This is why I never bought an autoloader - I couldn't figure out where
the autoloader commands came from. DEC might sell more autoloaders if
they would tell people how to use them.



Note 1482.26, 5-Jun-1995
Hassinger: 
-----------
Re .24 - thanks!
Re .25 - amen!!



Note 1482.27, 5-Jun-1995
Hassinger: MK and GK on the same unit???
----------------------------------------
> The answer to his question is that you use IO$_DIAGNOSE QIOs through the
> GKDRIVER to LUN 1 of the device ...
    
Question then.  As I mentioned, it would seem VMS normally sees the
tape part as MKannn.  I am not clear what, if anything, is involved in
trying to use GKDRIVER to access the autoloader.



Note 1482.28, 6-Jun-1995
Nagy: Canot have MK and GK on the same unit!
--------------------------------------------
Both the MKDRIVER and DKDRIVER provide support for sending your own
SCSI commands to the device.  You need DIAGNOSE privilege and then
can use the IO$_DIAGNOSE function code to send SCSI command packets






























The DECUServe Journal  July, 1995                              Page 43
4mm DAT Autoloader


to the device.  This appears to be the way the DLT stackers works
since the stacker mechanism is not a separate SCSI id.

This should also work when then stacker mechanism is a higher numbered
LUN (see above) on the same SCSI id.

> Question then.  As I mentioned, it would seem VMS normally sees the
> tape part as MKannn.  I am not clear what, if anything, is involved in
> trying to use GKDRIVER to access the autoloader.

One has to manually load the GKDRIVER and configure it at startup.
A particular SCSI id can only be serviced by a single driver so a
device handled by the DKDRIVER or MKDRIVER cannot be accessed by the
GKDRIVER (but, see above).  We have several Exabyte 10i stackers and
EXB120 8mm libraries.  The robot controllers are separate SCSI ids
and are accessed by the GKDRIVER.



Note 1482.29, 6-Jun-1995
Parris: 
--------
I forget exactly how SCSI unit numbers with LUNs are formed, but for a tape
drive MKA500, LUN 1 could be accessed through GKA501 or GKA510, I forget which.
You'd create the GKA device with a CONNECT command under SYSGEN (or SYSMAN IO
on Alpha). 



Note 1482.30, 8-Jun-1995
Hanrahan: 
----------
xxx501.  eg with the Pioneer CD-ROM changers you let it autoconfigure
LUN 0 as, for instance, DKA400, and then SYSGEN CONNECT DKA401, 402, 
_..., 405.  




  RRD50/KRQ50 Info
  ----------- ----

  Abstract:

  Following is a brief discussion of an old RRD50 CD-ROM drive and
  corresponding KRQ50 Q-Bus controller, with a few details on cables
  and DIP switch settings, and an offer of a manual.  (Now don't go
  bombarding poor Bruce with requests for it!  That's what the
  SWAP_MEET conference is for.)



































The DECUServe Journal  July, 1995                              Page 44
RRD50/KRQ50 Info


  Participants:

  Barton Bruce, Jeffrey Campbell.

  Conference:  HARDWARE_HELP


Note 1882.0, 4-Jun-1995
Campbell: RRD50/KRQ50 info request
----------------------------------
I have added an RRD50 CDROM drive and KRQ50 controller to my hardware
collection. Unfortunately, I have no manuals or inter-connecting cable.
I have figured out the KRQ50's addressing jumpers and VMS finds it.
But I don't have the cable between the controller and the drive. I can
make a cable if I have the pinouts on the drive DB-15 and the KRQ 20 pin
BERG.

I also am curious what the DIP switches on the rear of the RRD50 control.

Help! 8-)



Note 1882.1, 9-Jun-1995
Bruce: 
-------
Groan. Some cables and 2 dead drives went to the dumpster within the last
week. One BC18R-6 cable fell off in the parking lot and may have made it into
a copper wire recycle pile - I'll watch for it.

The ribbon cable to bulkhead kit is in my hand now. Barely readable
is P/N 70-21013-01. The ribbon cable splits in half with each half
going to a DB15 Female with the ribbon cable crimped into
the first 10 sockets. Wires 1-10 probably go to 1,9,2,10,3,11,4,12,5,13
of connector A and 11-20 the same ones of connector B, but I am guessing.

BC18R-6 cables go from the bulkhead to the drive.

SW-1 bottom is address bit 2, the top is bit 11. Not so obvious is that 
bit 12 is J7 just above the switch.

        +
                        = 0
        +-+


        +
        |               = 1
        + +


> I also am curious what the DIP switches on the rear of the RRD50 control.

2 manuals were in the pile to be tossed TODAY.






























The DECUServe Journal  July, 1995                              Page 45
RRD50/KRQ50 Info



The drive user's quide shows switch 4 (MSB) through 9 (LSB) counting
unit numbers from 0 (all off) to 16 (just 4 on).

What I have left, including the Q-Bus controller (if I can find it) are
yours if you want them.




  Wildcard Algorithm
  -------- ---------

  Abstract:

  A question in the rarely-used (but apparently closely-watched)
  ALGORITHMS conference concerning wildcard pattern-matching in C gets
  a quick response.



  Participants:

  Bruce Bowler, John Briggs, Dale Coy, David Mischler.

  Conference:  ALGORITHMS


Note 40.0, 18-Apr-1995
Bowler: Wildcard matching.
--------------------------
I'm looking for an algorithm or pointer to one for a fast, reliable
"does this string match this string that contains wildcards". 
Wildcards can, as in VMS match 1 (%) or any number (*) of characters.
    
ANSI C code would be most helpful :-)



Note 40.1, 18-Apr-1995
Mischler: Would full regular expressions be OK?
-----------------------------------------------
Have you considered looking at GNU grep?



Note 40.2, 18-Apr-1995
Coy: Grep isn't an algorithm
----------------------------
There are lots of "tools" that will do matching.  However, Bruce asked
for an algorithm.  
    
Bruce, if TOOLS will do, let us know.































The DECUServe Journal  July, 1995                              Page 46
Wildcard Algorithm


Note 40.3, 19-Apr-1995
Bowler: 
--------
Nope.  I need to implement some pattern matching code in a system I'm
working on.  It needs (or will eventually need) to run on several
systems.  Language of choice is C, only need %* matching, don't want
the "complexity" of grep.



Note 40.4, 19-Apr-1995
Briggs: C code for pattern match
--------------------------------
It's not that hard to do.  After fixing the typos, changing the main
routine's "return" to SYS$EXIT and using argv[1]/argv[2] instead of
argv[0]/argv[1] (C is not my language of choice), this code actually
worked right the first time.

At least on the test cases I threw at it.  Your mileage may vary.

main(int argc, char* argv[])
{
    int match ( char* pattern, char* string);
   
    sys$exit ( match(argv[1],argv[2]) );
}

int match ( char * pattern, char * string )
{
    int pi;
    int si;

    pi = 0;
    si = 0;
    while ( pattern[pi] != 0 ) {
        if ( pattern[pi] == '%' ) {
            /* Match exactly one arbitrary character */
            if ( string[si] == 0 ) return 0;
            pi++;
            si++;
        } else if ( pattern[pi] != '*' ) {
            /* Match one specific character */
            if ( string[si] != pattern[pi] ) return 0;
            pi++;
            si++;
        } else {
            /* Match arbitrary number of arbitrary characters... */
            /* All sub-patterns of "*" and "%" are equivalent to */
            /* a simpler pattern of a single * and the right number of %'s */
            /* That is, it matches an arbitrary sequence of _at least_ */
            /* some number of characters.  Skip that many characters. */
            while ( pattern[pi] == '*' ||
                    pattern[pi] == '%' ) {
                if ( pattern[pi] == '%' ) {






























The DECUServe Journal  July, 1995                              Page 47
Wildcard Algorithm


                    if ( string[si] == 0 ) return 0;
                    si++;
                }
                pi++;
            }
            /* The next character in the pattern is either a text character */
            /* or end-of-string */
            if ( pattern[pi] == 0 ) return 1;   /* Match remainder of string */

            /* If it's text, test all possible matches recursively */
            while ( string[si] != 0 ) {
                if ( string[si] == pattern[pi] ) {
                    if ( match(&pattern[pi+1],&string[si+1]) ) return 1;
                }
                si++;
            }
            /* None of them matched */
            return 0;
        }
    } /* End while */
    /* If we get here, the pattern ran out without encountering a "*" wildcard
*/
    /* If the string has also run out, it matches.  If not, it doesn't match */
    return ( string[si] == 0 );
}



Note 40.5, 19-Apr-1995
Bowler: Thanks!
---------------
    



  DTR column headers
  --- ------ -------

  Abstract:

  A discussion of how to control the printing of column headers in
  Datatrieve.  The default handling can, under sufficiently complex
  conditions, fail to perform as desired.  Options discussed include
  use of the Report Writer and explicitly constructing column heads.



  Participants:

  John Briggs, Bob Hassinger.

  Conference:  4GLS_AND_QUERY_TOOLS
































The DECUServe Journal  July, 1995                              Page 48
DTR column headers


Note 203.0, 19-Apr-1995
Hassinger: Controlling when DTR prints column headers?
------------------------------------------------------
This is an old issue, but I am not sure if I have ever seen the answer. 
It is about how to control when Datatrieve prints column headers.

Consider:

    FOR  whatever
        BEGIN
        PRINT something("col"/"title"),something_else("col2"/"title")
        END

DTR prints the column headers before the first line of output, and then
is clever enough to not do it again for each succeeding line printed. 
That is fine.

But now consider:

    WHILE something that selects new sets sets of data
        BEGIN
        PRINT NEW_PAGE
        FOR  whatever
                BEGIN
                PRINT something("col"/"title"),something_else("col2"/"title")
                END
        END

This still only prints the headings once, even though we select new
data sets, and go to new pages to report on each set.

I want to be able to convince DTR to print the column headings each
time I loop through the WHILE and start a new page.  So far I have
been unable to find a way to do it.

BTW - the real program is more complex of course - there is an ON
outside the WHILE so it all goes to a file for printing, there is
actually more than one FOR inside the WHILE with different print
commands, and a form is used in each WHILE loop to get the key that is
used in the FORs to select the data sets.

As it turns out, DTR seems to be able to track that it has printed the
column headers, even with multiple, different FOR/PRINTs in the loop. 
Not a bad trick...

But a pain.  Any insight into how this really works and/or how to
control it and be able to force the column headers to print on each new
iteration through the WHILE?




































The DECUServe Journal  July, 1995                              Page 49
DTR column headers


Note 203.1, 20-Apr-1995
Column header on nested print.  Solutions
-----------------------------------------
Try using the report writer for the inner print:

    WHILE something that selects new sets sets of data
        BEGIN
        PRINT NEW-PAGE
        REPORT whatever
                AT TOP OF REPORT PRINT COLUMN-HEADER ! But not page-header
                PRINT something("col"/"title"),something_else("col2"/"title")
        END-REPORT
        END

Of course, the report writer imposes its own restrictions about the kind
of actions you can perform within its scope.  Depending on your requirements,
you may be able to end-run around some of those restrictions.

Alternatively, you could construct your own column header.  Perhaps
something like:

DECLARE COLUMN-HEAD COMPUTED BY "GROSS    NET            " |
                                " PAY     PAY"
                    EDIT-STRING IS T(24).
PRINT NEW-PAGE, COLUMN-HEAD(-)

> As it turns out, DTR seems to be able to track that it has printed the
> column headers, even with multiple, different FOR/PRINTs in the loop. 
> Not a bad trick...

No big deal.  Each PRINT statement keeps its own context.  When the whole
nested mess exits back to command level, the context is discarded (along
with the statement itself).



Note 203.2, 20-Apr-1995
Hassinger: 
-----------
Thanks for the suggestions.

> Try using the report writer for the inner print:

I think back in 88 when I wrote this and last looked at it I was unable
to find a way to get the results I wanted with Report Writer.  One
problem may have been that there is  more than one FOR inside the
WHILE, each doing different things.  The first one reports several
columns from each record of a particular type, then the next reports a
number of different columns from another kind of records.  There is
also some special case logic that can print other things under selected
lines that have special situations and so on.

> Alternatively, you could construct your own column header.  Perhaps
> something like:






























The DECUServe Journal  July, 1995                              Page 50
DTR column headers


>
> DECLARE COLUMN-HEAD COMPUTED BY "GROSS    NET            " |
>                                " PAY     PAY"
>                    EDIT-STRING IS T(24).
> PRINT NEW-PAGE, COLUMN-HEAD(-)

I suspect that is more-or-less what I will have to do.  I guess it
means explicitly controlling column widths and placement so the header
and the data lines line up.

BTW - is there an advantage to doing it with the declaration like that
rather than with simple PRINTs (these headings actually end up down in
the page at the start of each of the different FOR loops, so they don't
come at the same point as the PRINT NEW-PAGE).



Note 203.3, 20-Apr-1995
Briggs: Sounds like you already know what you've got to do
----------------------------------------------------------
> I suspect that is more-or-less what I will have to do.  I guess it
> means explicitly controlling column widths and placement so the header
> and the data lines line up.

Yep.  Or you could fit the header to whatever layout the PRINT statement
generates by default.

> BTW - is there an advantage to doing it with the declaration like that
> rather than with simple PRINTs (these headings actually end up down in
> the page at the start of each of the different FOR loops, so they don't
> come at the same point as the PRINT NEW-PAGE).

No advantage that I can think of.  Unless you have one heading that fits
more than one print statement.




  Saving Window Positions
  ------ ------ ---------

  Abstract:

  Understatement:  Not everything about DECwindows is obvious to
  someone just starting out.  People accustomed to certain other
  windowing user interfaces (which will remain nameless, but they're
  not from MIT) might expect, for example, that windows will remember
  their positions from one session to the next.  Of course, things are
  rarely as simple as they seem.



































The DECUServe Journal  July, 1995                              Page 51
Saving Window Positions


  Participants:

  david Campen, Rick Carter, Larry Kilgallen, Brian Tillman, Don
  Vickers, Steve Wilkins.

  Conference:  USER_INTERFACES-WINDOWS


Note 217.0, 6-Jun-1995
Carter: Saving window positions?
--------------------------------
OK, I'm a 24-hour-new-to-DECWindows person.  I've set all my stuff to
startup automatically -- clock over here in one corner, MAIL over
there, all very lovely.  Now I want to get out for the day and come
back in tomorrow.  Everything doesn't stay in its lovely place.  Is
there anyway to save the window positions of the programs you set to
startup automatically?



Note 217.1, 6-Jun-1995
Kilgallen: 
-----------
Typically this is done by the individual application ... or not.



Note 217.2, 6-Jun-1995
Campen: 
--------
I just leave my session active. I use pause to keep other people out
and turn off the monitor. 



Note 217.3, 6-Jun-1995
Vickers: All I know about resource files (and maybe more :')
------------------------------------------------------------
Staying in the DECwindows session works especially if you failed to
learn how to share in kindergarten.  (':

My memory is hazier than normal on this as it has been several years
since I hacked around in DECwindows.

The various settings for DECwindows applications are in resources files
named DECW$.DAT.

These files are used if they exist in the SYS$LOGIN directory.  If they
are not there, the system searches through the DECW$SYSTEM_DEFAULTS
logical name.

Your first check should be to look for any DECW*.DAT; files in
SYS$LOGIN as some applications copy the system default file there as I
recall.  If the resource files of your favorite DECwindows applications






























The DECUServe Journal  July, 1995                              Page 52
Saving Window Positions


are not in SYS$LOGIN then get a directory of
DECW$SYSTEM_DEFAULTS:DECW*.DAT and copy the usual suspects from there.

In spite of the .DAT name, the resource files are just text files so
you can edit them to your heart's content.  The first resource in most
of the files is the geometry resource:

    Mail.geometry:  555x610+174+20
                     ^   ^   ^  ^
    Window X size ___+   |   |  |
    Window Y size -------+   |  |
    Window X location -------+  |
    Window Y location ----------+

If the geometry resource isn't there then you can add it.  You can
experiment with the X and Y values or observe the location values as
you drag the window around the screen.

In most of the resource files, there are usually enough comments to
give you ideas about what the other resources do.

Playing with resource files can provide hours of entertainment and kill
a lot of time.



Note 217.4, 7-Jun-1995
Wilkins: A DECW setup that works for me
---------------------------------------
I "stole" the following concept from here or some other system (the
memory goes first).  I have been using it successfully for years.
    
The first file is a command procedure that is executed each time I
fire up my workstation.  Note that I am not interested in location
for all my sessions.
    
    
$ create/terminal -
        /detach -
        /resource_file=sys$login:decw_term2.dat -
        /window_attributes = -
             (X_POSITION=5,Y_POSITION=467,initial_state=icon)
$ create/terminal -
        /detach -
        /resource_file=sys$login:decw_term1.dat -
        /window_attributes = -
             (initial_state=icon)
$ create/terminal -
        /detach -
        /resource_file=sys$login:decw_term5.dat -
        /window_attributes = -
             (initial_state=icon)
$ create/terminal -
        /detach -






























The DECUServe Journal  July, 1995                              Page 53
Saving Window Positions


        /resource_file=sys$login:decw_term3.dat -
        /window_attributes = - 
             (X_POSITION=75,Y_POSITION=295,initial_state=icon)
$ create/terminal -
        /detach -
        /resource_file=sys$login:decw_term6.dat -
        /window_attributes = -
             (initial_state=icon)
$ create/terminal -
        /detach -
        /resource_file=sys$login:decw_term4.dat -
        /window_attributes = -
             (X_POSITION=5,Y_POSITION=381,initial_state=icon)
    
==================================================================
    
An example of the decw_term*.dat file follows (nothing fancy - just
naming the windows and Icons).
    
==================================================================
    
DECW$TERMINAL.decWmValueMask:   1
DECW$TERMINAL.title:    RITVAX
DECW$TERMINAL.main.terminal.displayWidth:       907
DECW$TERMINAL.main.terminal.displayHeight:      488
DECW$TERMINAL.main.terminal.displayWidthInc:    11
DECW$TERMINAL.main.terminal.displayHeightInc:   20
DECW$TERMINAL.main.terminal.fontSetSelection:   0
DECW$TERMINAL.main.terminal.fontUsed:
   -Bitstream-Terminal-Medium-R-Normal--18-180-75-75-C-11-ISO8859-1
DECW$TERMINAL.main.terminal.bitPlanes:  4
DECW$TERMINAL.main.terminal.y:  20
DECW$TERMINAL.main.terminal.width:      907
DECW$TERMINAL.main.terminal.height:     488
DECW$TERMINAL.main.terminal.borderColor:        #111100001111
DECW$TERMINAL.iconName: RITVAX
DECW$TERMINAL.borderWidth:      1
    
==================================================================
    
Believe me you can have hours of "fun" locating windows on a screen.
    
Hope this helps.
    


Note 217.5, 15-Jun-1995
Tillman: Where to find Save
---------------------------
After positioning your clock, put the pointer on its face, press and
hold MB3, and you'll get a pop-up menu one of whose items is "Save
Settings".  Slide the pointer to it and let go.  The next time, your
clock will be in the same location.  You can do this for Mail, too, but
the "Save Options" menu item is in the "Options" menu of the main






























The DECUServe Journal  July, 1995                              Page 54
Saving Window Positions


window.  The same for DECterms, CMS, LSE, and some others.  The
Calculator has a "Save Geometry" item in the File menu.
    
This is not an exhaustive list.




  About the DECUServe Journal
  ---------------------------

  Publication Information

  Topic threads in the DEC Notes conferences on DECUServe are selected
  for publication on the basis of strong technical content and/or
  interest to a wide audience.  They are submitted to the editor from
  various sources, including DECUServe Moderators, Executive Committee
  members, and other volunteers.  Suggestions for inclusion are
  enthusiastically solicited.  Articles selected for publication are
  edited on an OpenVMS VAX system in TPU and then formatted with
  Digital Standard Runoff.

  What's a DECUServe, Anyway?

  DECUServe is an on-line conferencing system which is available 24
  hours a day, seven days a week.  System backups are done on Fridays
  from 7:00am to approximately 9:00am Eastern (Daylight/Standard)
  Time.  Scheduled downtimes for equipment or software maintenance are
  announced in advance.  The system can be reached worldwide, and
  membership is by individual subscription only (no group or company
  accounts).  The current annual subscription fee is US$75.00.

  On-line subscription information is available in the U.S.  by
  dialing 1-800-521-8950 and logging in with username INFORMATION.
  DECUServe and the INFORMATION account can also be reached on the
  Internet via telnet connection to decuserve.decus.org or e-mail to
  information@decuserve.decus.org.



  Contact Information
  -------------------

  The editors of the DECUServe Journal are Brian and Sherrie McMahon.
  They can be reached by any of the following means:

       mcmahon_b@decuserve.decus.org
       mcmahon_s@decuserve.decus.org
       mcmahonb@decus.org
       griffith@decus.org
       mcmahon@ac.grin.edu

































The DECUServe Journal  July, 1995                              Page 55
Saving Window Positions


       griffith@ac.grin.edu
       +1 515 269 4901 (normal office hours, U.S. Central time)
       +1 515 269 4936 FAX