Article 7389 of comp.org.decus: Path: nntpd.lkg.dec.com!crl.dec.com!crl.dec.com!bloom-beacon.mit.edu!newsfeed.internetmci.com!tank.news.pipex.net!pipex!in2.uu.net!spcuna!eisner!mcmahon_b From: mcmahon_b@eisner.decus.org (Brian McMahon, Info-VAX Refugee) Newsgroups: comp.org.decus,vmsnet.decus.journal Subject: DECUServe Journal August 1995 Message-ID: <1995Sep2.144926.11116@eisner> Date: 2 Sep 95 14:49:26 -0400 Organization: DECUServe Approved: mcmahon_b@eisner.decus.org Lines: 2986 Xref: nntpd.lkg.dec.com comp.org.decus:7389 vmsnet.decus.journal:10 The DECUServe Journal --------------------- August, 1995 From the Editors' Keyboard . . . . . . . . . . . . . 2 What's Inside, Words to and from the Weary RFL the FLB . . . . . . . . . . . . . . . . . . . . 3 ALL-IN-1 Forms Library Management Tips Retracting a Mail Message . . . . . . . . . . . . . 6 Retracting Mail in ALL-IN-1 and Elsewhere Ethernet Interconnections . . . . . . . . . . . . 21 Hubs, Repeater Rules, Book Recommendation RZ26-VA Transplant . . . . . . . . . . . . . . . . 24 Moving a Drive to a 3100 Vanity Domains in URLs . . . . . . . . . . . . . . 26 Your Company Name Here -- How Does it Work? Article title goes here . . . . . . . . . . . . . 28 One-line summary goes here Serving CDs with PATHWORKS . . . . . . . . . . . . 30 Serving InfoServer CDs to PATHWORKS Clients PATHWORKS-to-PATHWORKS Copy . . . . . . . . . . . 32 Moving Docs from OpenVMS to NT PATHWORKS InfoServer CDs to Windows NT . . . . . . . . . . . 34 Another Look at Serving CD-ROMs to NT Clients Intel PC Booting . . . . . . . . . . . . . . . . . 35 Where Does an Intel PC Boot from, and When? DCL Error Trapping . . . . . . . . . . . . . . . . 40 A Common Misunderstanding of ON ERROR Low Priority Batch Jobs . . . . . . . . . . . . . 43 Tuning for Low Impact Batch Jobs WWW Advertising . . . . . . . . . . . . . . . . . 49 Concerns About Advertising on the Web The DECUServe Journal August, 1995 Page 2 From the Editors' Keyboard From the Editors' Keyboard ---- --- -------- -------- We hope you'll pardon us for being tardy (and a bit surly maybe, too) once again. August has been a "feature-rich" month. Those of you who work at academic institutions may know what we mean. But in spite of Midwest heat, bouts with illness, the start of another semester, and similar calamities, the August issue of the DECUServe Journal is here, and not a moment too soon. Such things being cyclical, we know that our schedule will settle down as we move on into September. It is even possible that this scorching Iowa summer will abate somewhat. Meanwhile, we thank you for your patience. If you see any spare rainclouds, or thirty-hour days, send them over this way. * * * * * The DECUServe Journal August, 1995 Page 3 RFL the FLB RFL the FLB --- --- --- Abstract: A discussion of various issues related to .FLB files in ALL-IN-1, including a nifty way to fix hard-coded file locations, and the hows and wherefores of forms library maintenance. Participants: Roger Bruner, Gary Burke, Bruce Burson, Lynda Peach, Don Vickers. Conference: ALL-IN-1 Note 1133.0, 14-Jun-1995 Bruner: Leftover Named Data in .FLB? ------------------------------------ For once I am writing about a solution rather than a question or problem. I had been locating a number of forms which somehow had sneaked in with a hardcoded file location rather than logical (Ugh!) and changing them to point to an appropriate logical. But it was hard to be sure that I had found all the forms. After seemingly finishing the job, I did a $ SEARCH OA$LIB:SITEOAFORM.FLB user$16 but I was still getting an awful lot of hits. I finally copied SITEOAFORM to my account and used Dale Coy's WPE (WPS-plus Emulator) editor to edit the .flb, where I searched for 'user$16'. Though I was obviously trampling through an area not intended for trampling through, it appeared that all of my hits related to all the forms I had changed. And, sure enough, if I did a search for the logical, it took me ALSO to the forms I had changed. I started wondering whether FMS retains old Named Data and, if so, how to get rid of it. Then, remembering how compact an .FLB is after doing a $ fms/lib/create siteoaform.flb siteoaform.flb I gave it a try and, sure enough, a search for 'user$16' revealed nothing! Now, to some of you, this may not be a surprise, but just in case this helps someone, I thought I'd share this. I was really of a mind to tear out my hair, and as many of you know, I haven't much to spare! The DECUServe Journal August, 1995 Page 4 RFL the FLB Note 1133.1, 14-Jun-1995 Burke: Last Things ------------------ Thanks for the interesting discovery, Roger. Isn't it funny how it's always the last thing you try that fixes the problem... Note 1133.2, 14-Jun-1995 Burson: Would RFL do the same thing? ------------------------------------ Note 1133.3, 14-Jun-1995 Vickers: Compression is your friend ----------------------------------- When I first read 1133.2 I was thinking of the CM RFL (Remove from live) option so I was about to get confused (again). I am sure that Bruce refers to SM M RFL (Reorganize forms library) which does do exactly what Roger did but with less typing and stuff. It is always a nice idea to keep your forms libraries compressed. FMS does not do this automatically. When you replace a form in the library, the old form is still taking up space in the library. It is just not available. The only way to eliminate the old wasted space is to compress the library. CM provides you with the ability to have the forms libraries compressed on a routine basis based on the number of updates. Unfortunately, many sites do not use the LFL option on the same menu to enter their site created libraries. I have seen a number of cases where sites have forms libraries of many thousands of blocks in size which went down to hundreds of blocks after being compressed. One site had a forms library taking up 10,000 blocks of space containing three forms (one named DUMMY, after me, I guess ;'). The library required 68 blocks after the compress. Compressing the forms libraries helps reduce both production and development times in addition to saving disk space. Note 1133.4, 15-Jun-1995 Bruner: Aha! ------------ I didn't realize about the routine compression of form libraries, but I have made a point for several years of periodically compressing the form libraries. I guess what this discovery yesterday helped reveal is what the compression really does! The DECUServe Journal August, 1995 Page 5 RFL the FLB I will pass along to my supervisor the information you provided, Don (BTW, how did you get DEC to name a form after you like that, Don? :-) Note 1133.5, 15-Jun-1995 Peach: When can you FDL safely? ------------------------------- Since you brought up the DUMMY.... can I ask a DUMB question? Can you run RFL while users are online? The "HELP" states: Do not reorganize the form library when it is being used by other people. If you try to reorganize a form library that is being used, any changes being made to it could be lost. The documentation states (p. 20-14 of Mgt Guide if anyone cares): Do not reorganize a form library at a time when users will be using it. If a form library is being used when it is reorganized, it is possible that changes made to the library by the programmer will be lost. Pretty much the same thing. I tend to read the word users are "customers" -- not people like me who might be working in the form library. Now, If I set up the system so that form libraries are automatically reorganized -- what does that do to the above? Looks to me like it would be automatically invoked while "users/customers" were using it... for example SITEOAFORM.FLB. Therefore, I suspect that both the HELP and documentation mean "programmers" instead of users here. Confirmation???? Note 1133.6, 16-Jun-1995 Vickers: RFL has a 'window' and programmers are people, too ----------------------------------------------------------- It does appear that the RFL does open a window of 'opportunity' for having a form placed into version n -1 rather than version n of the forms library. I looked through the code in the command procedure and it takes out a lock on the library but only while making a temporary copy for the reorganization. Once the actual work is being performed, someone updating the forms library would be using the old library rather than the new one. The 'someone' would have to be either a programmer or a maintainer. Programmers can affect the DEVELOP libraries and maintainers can affect the production libraries such as SITEOAFORM. This means that the use of the term 'being used by other people' refers to programmers and maintainers. I admit this usage is a bit strange but don't forget that the 'users' of CM are these very 'people'. The biggest exposure is PROBABLY to the DEVELOP libraries since these are being 'hit' by multiple programmers (if you have multiple active programmers). The DECUServe Journal August, 1995 Page 6 RFL the FLB What seems REALLY strange to me is that the regularly scheduled automatic reorganizations do not expose the library to this window because it does the action while the lock is being held. The bottom line is that the documentation is correct and it does mean programmers and 'worse'. Note 1133.7, 16-Jun-1995 Peach: Good news - thanks ------------------------- Thanks a lot Don -- that's "good" news for me since it means I don't have to figure out a good time to do this and have ALL-IN-1 shutdown. Retracting a Mail Message ---------- - ---- ------- Abstract: How (and whether) one might retract a message after sending it. Although parts of the discussion are technical and highly specific to ALL-IN-1, broader issues come up as well. Is such a feature even desirable? How might it best be implemented? Participants: Roger Bruner, Dale Coy, Lynda Peach, Graham Pye, Don Vickers, Ray Whitmer, Anonymous Canadian. Conference: ALL-IN-1 Note 1123.0, 30-May-1995 Bruner: "Cancelling" mail messages ---------------------------------- Today marked the third time in the past six months that I have received a frantic call from a fairly high-level User who has sent a message before intending to or sent it to the wrong people or attached the wrong attachments, etc. Apparently there has been a 100% genuine need to retrieve those messages -- preferably before they were read, but at least before they were read by very many addressees. I developed a simple procedure the first time which I have kept refining, but the further I go, the more questions I have. I first NEWDIR to the sender's account while I have him/her on the phone to identify the message. I select it and run my System The DECUServe Journal August, 1995 Page 7 Retracting a Mail Message UDP, UNSEND. UNSEND prompts me whether to operate on a message or a filename; I answer 'message'. Then it asks whether to use the current message; I answer 'yes'. The UDP then retrieves the actual filename [from the appropriate shared area] and copies to that filename a file that basically says, "The original message was sent by mistake. The Sender has asked that this message be used in its place." Something like that. And that so relieves the Sender that we could almost leave it at that. However, today's problem message had an attachment, and the attachment was as sensitive as the message itself and had to be dealt with separately. I started to modify my UDP to include something like (note that I am omitting the UDP directives and giving only the DO SCRIPT syntax here): for cab$attach do get #fil = .value\\- copy "oa$lib:mis$sent2.wpl" #fil and that was okay. "MIS$SENT2.WPL" says something like, "This takes the place of an erroneously sent attachment." Then I started testing this UDP (long after the current need was taken care of, I hasten to add), only to discover that my FOR LOOP was actually wiping out the REAL text of all attachments that were mail messages. I have modified the FOR LOOP something like this: for cab$attach with .dtype nes "mail" do - get #fil = .value\\- copy "oa$lib:mis$sent.wpl" #fil and am simply leaving alone messages that are attachments. I can understand why documents are safe, because they (copies, that is) are placed in the Shared Area as soon as attached. Whereas mail messages are already unmodifiable after being sent and are already in the Shared Area. But I feel I need to operate on messages, too, or I may still fail someone in his/her hour of need. Are any of you dealing with similar problems? Am I trying to stand on my head unnecessarily? What is/would be your method, and/or what are your suggestions for improving mine? The DECUServe Journal August, 1995 Page 8 Retracting a Mail Message Note 1123.1, 5-Jun-1995 Pye: Some thoughts ------------------ First answer is to charge them $500 each time you do it, and then they'll soon stop asking you :-) I think a simple solution is to delete the SDAF record, and then re-write it pointing to a copy of your "it's gone away text" but without any of the attachment pointers etc. This would mean that the attachments would end up with usage counts that were too high, but FCVR will fix that later. You'd need to give your dummy SDAF record a usage count that was the same as the original, so that it wasn't deleted until everyone had read it. I'm not sure how much of this is doable without *real* code though... Note 1123.2, 5-Jun-1995 Bruner: The FCV what? --------------------- That sounds workable, Graham, except for the need for "real code"! BTW, one drawback to my being in a support position (rather than A1 Manager) is that we NEVER run the FCVR. I have mentioned it periodically...enough said here. Note 1123.3, 5-Jun-1995 Peach: Think TRM ---------------- What do _they_ run instead? FCVR or TRM is, imho, vital to the health of an ALL-IN-1 system. I can't believe they are willing to go through these hoops to help a user out in Email and fail to run something that COMES WITH THE SYSTEM that is just as vital to the health of Email. Note 1123.4, 6-Jun-1995 Bruner: Agreed, but... ---------------------- I agree, Lynda. I think part of the problem goes back several versions of ALL-IN-1 to a time when the FCVR was less robust. I hesitate to say any more than that, as you can well appreciate. I'm sure I'd be in trouble if it became known that I'd been discussing this "publicly". Note 1123.5, 7-Jun-1995 Pye: On the other hand... ------------------------- Re .1 The DECUServe Journal August, 1995 Page 9 Retracting a Mail Message I thought about this some more, and you'd need to look out for SDAF continuation records too. Rather than my cavalier idea of letting the FCVR fix up the usage counts you probably need to be more subtle. Forget I spoke :-) Note 1123.6, 7-Jun-1995 Bruner: Don Vickers' fault for sure! ------------------------------------ It's easy to think things are easier than they are, isn't it, Graham? Actually, though I would have loved a nice simple solution, it has been reassuring that at least no one came back and said, "Why didn't you use the xxx option on the yyy menu? It's just for that purpose!" In a sense, any success at "cancelling" a message is still some help to the User; after all, it was the User's error, not mine, not ALL-IN-1's. But -- for better or worse -- somewhere along the line I got bit by Don Vickers' concept of excellence in terms of really trying to help the Users, and I believe what I want to do is based on that disease! Note 1123.7, 8-Jun-1995 Whitmer: Has this feature been requested? ----------------------------------------- Has this been requested as an ALL-IN-1 feature. I think of how often and easily I retract mistaken mail messages or other objects across a diverse WAN using GroupWise, and how useful it is, and have a hard time relegating this situation purely to user error where the user is lucky to get anything! Or maybe I am missing something? disclaimer: I have a vested interest in GroupWise as I work for Novell. I just think that a feature as useful as this should have become standard functionality in any package long ago (we added it for the Department of Justice about ten years ago, and it is still heavily used and highly acclaimed). Note 1123.8, 8-Jun-1995 Bruner: Requested...how, anymore? --------------------------------- Ray, IMHO the ability to request that such things be added to the base product has gone away. Whereas we know Graham is alive and listening and -- yes! -- concerned about our opinions, and whereas I am sure there are still some of the Reading group who would love to help, in the absence of a WISHLIST process, I just don't feel good about the possibility of our requests going too far (though I'll give an 'A' to the good intentions of everyone involved). How's everyone else feel? The DECUServe Journal August, 1995 Page 10 Retracting a Mail Message P.S. I got called again today to use my cancellation procedure. Maybe I'd better tell my supervisor about it; he may want/need to use it himself, too! Note 1123.9, 8-Jun-1995 Coy: I would oppose a "retract" function for users -------------------------------------------------- It's a question of philosophy and "user paradigm". And implementations certainly disagree on those two things. Philosophy -- ALL-IN-1 takes the viewpoint that e-mail is best modeled on the postoffice system. Once you put it in the mail, you can't get it back, and you can "expect" that it will be delivered (unless you hear otherwise). And "expect" doesn't mean perfection. User paradigm -- for a "retract" function, the "normal" user would expect that the mail would not be delivered at all (nobody could have seen it). The "above-average" user would expect that someone might have received and read it, but that all other deliveries would be destroyed before they can be read. BUT, you and I know that's not possible. As a simple example, forwarded mail will probably not be retracted, etc. My _bet_ is that GroupWise isn't perfect -- that is, that it can't always retract mistaken mail messages. How do you explain the meaning of "retract" to normal users? If someone retracts a mail message, how does he/she determine who might have read it, and who will not be able to, and who might still be able to read it later (e.g., it's been delivered where it can't be retracted)? That's not a criticism -- it's just an alternative view that says "it doesn't make sense to implement something imperfectly". An equally valid view would be "better imperfect than not at all" (which is apparently the GroupWise view). If there were a wishlist, I would not support a built-in user function in ALL-IN-1 to "retract e-mail". However, I would support an "administrator" function. But that's just one person's view. Note 1123.10, 9-Jun-1995 Bruner: FMB mail different from U.S. mail! ------------------------------------------ Dale, what you say about the post office model makes sense. I might add that at the Foreign Mission Board we have a mailbox of our own near the snack room which gets emptied out and taken to the actual post office (by our staff) at 5:00 or so; more than once, I have seen one of our mail-handling folks opening that up to retrieve a letter unintentionally dropped in. So I guess you'd have to say we tend to model our ALL-IN-1 mail after our FMB mail system! The DECUServe Journal August, 1995 Page 11 Retracting a Mail Message Note 1123.11, 9-Jun-1995 Whitmer: You are right. The feature would not fit in isolation. ---------------------------------------------------------------- Dale, You are absolutely correct that it can be impossible in certain circumstances to retract mail. And GroupWise refuses to retract from a user has already read or forwarded it. As you have implied, just because a function may not be possible in all cases, it should not be offered. If that were true, then we couldn't offer mail delivery in the first place. ALL-IN-1 may typically only execute in a more homogeneous predictable environment than GroupWise. The key, as you have identified it, is whether the user easily knows when and where it failed. Groupwise has an in box and an out box. The outbox contains the mail a user has sent. By clicking on any piece of mail in the outbox, a user is shown the status of that message with respect to every recipient: Who has received the message on their system, who has read the message, who has deleted it, who deleted it without reading it, etc. Also, when each of these operations happened. It's kind of like having UPS who says they can tell you exactly where each piece of mail is at any time. When mail is retracted, this status is updated to show exactly if/when each deliveries were successfully retracted, as well as mail which was already read or went onto systems which do not support retraction, etc. But really most of the time retraction is as reliable as the original delivery. But if the retraction was important, everyone looks to see whether it was successful, the same as they look to see if mail they sent was delivered or read. > That's not a criticism -- it's just an alternative view that says "it > doesn't make sense to implement something imperfectly". An equally > valid view would be "better imperfect than not at all" (which is > apparently the GroupWise view). > > If there were a wishlist, I would not support a built-in user function > in ALL-IN-1 to "retract e-mail". However, I would support an > "administrator" function. But that's just one person's view. I suspect that without a good up-to-date status report on your mail as I have described, such functionality would be much less useful. I clearly oversimplified by only considering the retraction functionality in isolation instead as a part of the greater functionality and paradigm that makes it useful. Oh well. It was worth a suggestion. The DECUServe Journal August, 1995 Page 12 Retracting a Mail Message Note 1123.12, 9-Jun-1995 Coy: Very interesting --------------------- Yep, it's good to discuss these differences in mail paradigm. > As you have implied, just because a function may not be possible in all > cases, it should not be offered. If that were true, then we couldn't I think I meant that the "reliability and predictability" is one of the things that enters into the tradeoffs. > Groupwise has an in box and an out box. The outbox contains the mail a > user has sent. By clicking on any piece of mail in the outbox, a user > is shown the status of that message with respect to every recipient: How do you handle "off system" mail? For instance, if mail is forwarded via SMTP, what do you show? Or, in the abstract, if it was delivered to some "dumb" mail agent (say VMSmail), what do you show? > Who has received the message on their system, who has read the message, > who has deleted it, who deleted it without reading it, etc. Also, when Interesting. 1. How do you determine that someone has read the message? [:-) I presume you really mean "opened"] And does it report whether the message was opened by the addressee, or by someone else (proxy)? 2. How would your mail system handle someone like me who objects strongly to the sender knowing whether I deleted it (or deleted it without reading it)? Not even UPS provides such a service. It would be extremely offensive to me. Anyhow, I certainly have heard the request for retraction capability enough times, as well as many other such requests. And I'm not saying there's anything "wrong" with it. These capabilities and implementations are what differentiates mail systems. As illustrated above, a feature that some people strongly desire can also be very offensive to other people. [Other easy examples are "blind carbon copy" and "forced read receipt"]. I find it very instructive to look at these different design choices made by different vendors. WRT "retraction", I agree (and never disputed) that it is useful in some situations. Obviously it adds complexity to the software, and I personally believe that it significantly adds to user training and "help desk" work. This is an illustration of the "no free lunch" principle, and both designers and purchasers have to decide whether the capability is worth the net cost. The DECUServe Journal August, 1995 Page 13 Retracting a Mail Message Note 1123.13, 9-Jun-1995 Vickers: A few random comments ------------------------------ Roger has already blamed me for his situation. I am sure someone else will come along and point out that I have only half a brain as well. Roger has done an excellent job of providing value to his clients and it works in HIS environment. I suggest that he should always warn his clients that even in his relatively simple environment, there is no way to know whether the receivers have received the retracted information or not. Since there seems to be a recurring need for this retraction, there may be some need for client training or counseling. About ten years ago I tried to make the case to the fine folks in engineering that retraction on a single system like Roger's still adds value. I tried to use an old "I Love Lucy" episode to describe how one can retract mail from the postal system using 'freshly used' chewing gum and a string. The English were spared that particular piece of Americana. My example failed as did I in convincing them that adding the feature was not a horrible idea. Their point at the time that a feature that works 'most of the time' is not reliable and therefore not providing 'service' to the clients. I have no knowledge of GroupWise and really prefer to not get into a weenie pull about competitive products. I do question whether Ray is correct in saying "ALL-IN-1 may typically only execute in a more homogeneous predictable environment than GroupWise." ALL-IN-1 is STILL used by many very large corporations as their corporate mail delivery system interfacing with a wide range of other systems. It has proven itself to be highly reliable in this world. The designers of the ALL-IN-1 system placed a very high premium on reliability and, IMHO, did a great job of it. As a number of companies have discovered in the past year or so, many of the more 'modern' products such as ccMail and MS Mail did not concern themselves as much with reliable service and large scale operations. Note 1123.14, 9-Jun-1995 Bruner: A little of this, that, & the other ------------------------------------------- Don, I didn't "see" you smiling at my concept of "blame" for making me more user-conscious. I wish I could be blamed for something equally terrible that I did to help someone else, especially in terms of attitude more than just the accomplishment of a single task. Yes, the mail retraction capability (thanks, Dale [I think], for giving this such a dandy handle) does involve a training factor (avoiding the need, i.e.). But the further we go with this, the more aware I am that the very fact of having ANY type of retraction capability COULD be taken as an unintentional invitation to additional carelessness on the part of the Sender ("After all, the ALL-IN-1 management section can always retract this."). Of course, there's no guarantee it hasn't been read by everyone first! The DECUServe Journal August, 1995 Page 14 Retracting a Mail Message Don, I sure appreciate knowing that you TRIED to get this functionality during your Digital years. Knowing you as I do, I'm not at all surprised that you did and that you probably felt pretty frustrated at the response you got to your suggestion. Ah, well... Note 1123.15, 9-Jun-1995 Whitmer: We are trying to support a business, rather than a feature. -------------------------------------------------------------------- I hope we are not really discussing this product versus that, but instead business processes and how they can be supported. I hope we are all over whose product has the most features. Simplicity, reliability, etc. count for much more. E-mail accounts for personal use are perhaps not much different than US Mail. But, in a large corporation, an E-Mail backbone must be much more. It can drive a large corporation allowing for asynchronous and optimal scheduling of all interactions that need to occur. I view it very much as the event queue of the corporation. It enables the coordination of many very complex business processes involving many people in very complex sets of processes and schedules. You get appropriate status when the message successfully leaves the statusable mail domain. Otherwise, I believe, you can track the progress of the message across servers until it is acted upon. With automatic rule-based message processing (autoforward is a special case) you can cause actions to appear to happen to your hearts content. Just like you can hack around in a computer program's event queue and refuse to conform and produce minor or major behavioral aberrations in the functionality of specific objects or the program as a whole. Managers who may spend 50% + of their time in E-Mail are not likely to be pleased. It is really important to know whether someone has had a chance to review and consider something. The rationale for each type of status seems to be founded in support for the business process, which could also be manually satisfied by reply mail, perhaps less efficiently (that is the ongoing debate, automation versus manual versus privacy). What is the impact in the case of retraction of the distraction of the original imperfect, outdated, incomplete, inappropriate message which is read by someone with any correction coming in a separate unconnected message. What is the impact of waiting for a week for someone to formulate a reply and not knowing that he hasn't been able to be in his office all week. Yes, there are other solutions without providing retraction. The constant input we get from user and business process research is that any action must be "forgivable". That's where the "Undo" feature idea comes from which is becoming universal. The user will never trust the program or himself with its power unless he can undo virtually any damage he might do. That is only one perspective. The DECUServe Journal August, 1995 Page 15 Retracting a Mail Message Note 1123.16, 9-Jun-1995 Coy: Not quite so simplistic ---------------------------- > I hope we are not really discussing this product versus that, but > instead business processes and how they can be supported. I hope we > are all over whose product has the most features. I'm not interested in "product wars" either -- and the issue is _never_ "feature count" (except when the marketing types print the brochures on slick paper...) When I did this kind of stuff for IBM, we all realized that there are different business processes, different corporate cultures, and different user orientations. It is in fact impossible for a single product (even with "build options") to do "best support" for all corporations (or all users). Some features actually conflict -- if you want this one, you can't have that one. You also have to watch out for terminologies that might _seem_ to be the same thing, but aren't. Example: IBM's idea of a "read receipt" is different from most other mail systems. The _most_ important real-user factor in e-mail systems is the "baby duck" factor. > The constant input we get from user and business process > research is that any action must be "forgivable". That's where the > "Undo" feature idea comes from which is becoming universal. Yes, I fully understand. But that's not exactly the subject we are discussing. To get a good assessment, you also have to ask questions like: o If we can't give you "forgivable mail send", does that mean that you won't by e-mail? (the black-or-white view). o If we can only retract mail under some situations, but not under others, how much value does that add? (shades of gray). If "retract mail" were such a *MUST HAVE* feature, then nobody would be using SMTP (or ALL-IN-1 or PROFS or SNADS or ...). I've even found instances where good research showed that users and managers "must" have a particular feature and -- when it was implemented as they demanded -- they demanded that the feature be removed. Note 1123.17, 9-Jun-1995 Bruner: A Solution that Seems to Work! -------------------------------------- BTW, a friend who's gotten shy about NOTE-ing sent me a nice mail message with some really useful suggestions for my retraction The DECUServe Journal August, 1995 Page 16 Retracting a Mail Message procedure. While they didn't work without some drastic modifications, they did get me moving in the right direction, so let me thank him AND share the results for anyone else who might want to know. This is in regard to the attachments question. Do a WRITE CHANGE CAB$ %KEY = OA$CURDOC, MODIFY="Y",MAIL_STATUS="UNSENT" and a similar line for DOCDB. I don't know for sure if the DOCDB line is necessary, but the CAB$ one sure is. Then I was able to do a MAIL DETACH until there are no attachments left. I'm sure there are other methods (I started with a FOR CAB$ATTACH DO MAIL DETACH, but that detached only 2 out of 3 attachments), but I ended up building an OA$TABLE:#TBL from 1 through the actual number of attachments (FOR CAB$ATTACH DO\GET #ATTS=OA$SEL_COUNT) and then FOR OA$TABLE:#TBL DO MAIL DETACH. This still presents me with an index of attachments which decreases by one with each one detached, but at least it did 'em all. Finally, I WRITE CHANGE my datasets back to SENT & UNMODIFIABLE. To test, I created a message with 3 attachments and addressed it to two other people (both are accounts I have access to). Then from the MANAGER account I NEWDIRed back to my account, selected the OUTBOX message and ran the procedure described above. Then I went into each recipient account and read the unread message; in both cases, there was no evidence that any attachments (let alone 3) had ever existed. I will test this also with situations where the message had already been read, but I can't see why the results should differ there. Of course, if READ already, a copy could have been made. Whattaya think? Note 1123.18, 10-Jun-1995 Vickers: A clever idea - a couple of improvements ------------------------------------------------- I wonder who could have given you that clever idea. ('; Any solution that does what you and your clients need is good and this one fits. First, the changing of the DOCDB is no longer required. In V2.mumble it was required but not now. Secondly, I suggest that rather than using MAIL DETACH that you would get faster and less visible service from CABINET DETACH_DOCUMENT. I have over engineered a script that uses the approach you and silent The DECUServe Journal August, 1995 Page 17 Retracting a Mail Message friend have developed. I am sure that someone with a full brain could do far better than a half wit like myself, and I will post it so the real experts can correct it. (': Note 1123.19, 10-Jun-1995 Vickers: ATTACHMENTS_STRIP.SCP ------------------------------ !+ ! ATTACHMENTS_STRIP.SCP Strip the attachments from a sent mail message ! ! This script was inspired by and for Roger Bruner by some anonymous Canadian ! person. ! ! It is invoked on the current mail message and removes all attachments so that ! receivers are not 'bothered' by them. ! !+ _.LABEL START ! DISPLAY (OA$_GBL_WORKING) FORCE GET #RB_GOOD_COUNT = #RB_BAD_COUNT = 0 ! ! Do a moderately rude thing - allow changes to the message ! GET #RB_MODIFY = OA$CURDOC_MODIFY GET #RB_MAIL_STATUS = OA$CURDOC_MAIL_STATUS ! .IF #RB_MODIFY NES OA$Y OR #RB_MAIL_STATUS NES "UNSENT" .THEN WRITE CHANGE CAB$ %KEY=OA$CURDOC, - MODIFY="Y", - MAIL_STATUS="UNSENT" .END_IF ! ! Determine the attachment names ! GET #RB_FILE = "" FOR CAB$ATTACH DO GET #RB_FILE = #RB_FILE .VALUE "," .IF OA$SEL_COUNT EQ 0 - THEN - .GOTO FINI ! ! Quietly detach the attachments ! FOR OA$TABLE:#RB_FILE DO - GET #RB_DETACH = .%WHOLE \\- .IF #RB_DETACH NES "" - THEN - CABINET DETACH_DOCUMENT , #RB_DETACH \\\\- .IF OA$STATUS EQ 1 - THEN - The DECUServe Journal August, 1995 Page 18 Retracting a Mail Message INCREMENT (#RB_GOOD_COUNT) - ELSE - INCREMENT (#RB_BAD_COUNT) ! _.LABEL FINI ! ! Restore the original CAB$ attributes ! .IF #RB_MODIFY NES OA$Y OR #RB_MAIL_STATUS NES "UNSENT" .THEN WRITE CHANGE CAB$ %KEY=OA$CURDOC, - MODIFY=#RB_MODIFY, - MAIL_STATUS=#RB_MAIL_STATUS .END_IF ! ! Report the actions and results ! .IF #RB_BAD_COUNT EQ 0 .THEN DISPLAY (#RB_GOOD_COUNT " attachments detached") .ELSE DISPLAY (#RB_GOOD_COUNT " detached & " #RB_BAD_COUNT - " failed - Press GOLD W") .END_IF ! .EXIT ! ! Author: Don Vickers, 904-326-2873 ! ! Copyright © 1995, Excellent Office Systems of Florida, Inc. ! ! This software may be copied and distributed freely as long as this statement ! of author and ownership is included. If you purchased this software from us, ! you are supported by our standard lifetime warranty against any defects in ! design or implementation. ! ! As always, be Excellent to one another and seize the day! ! ! Modified by: | Date: | Reason: !----------------------------------------------------------------------------- ! Don Vickers | 09-Jun-1995 11:40pm | Element created, please state ! | | modifications below ! | | ! ! [End ATTACHMENTS_STRIP.SCP] Note 1123.20, 10-Jun-1995 Whitmer: I'll have to admit my ignorance... ------------------------------------------- > -< Not quite so simplistic >- I didn't think I said it was. The DECUServe Journal August, 1995 Page 19 Retracting a Mail Message > The _most_ important real-user factor in e-mail systems is the "baby > duck" factor. I'll have to admit my ignorance, what is... Note 1123.21, 11-Jun-1995 Coy: Imprinting --------------- Oh. Alias "imprinting". The first e-mail system that you encounter is the one that you are "imprinted" on. For the rest of your life, you compare all e-mail systems to that one, and you "know" that it's features were the right ones. Note 1123.22, 12-Jun-1995 Bruner: Quackers? A religious group? ------------------------------------- Dale, I guess the "baby duck factor" does sound nicer than "the basis for lifelong prejudices"! I'm glad the question was asked; I've heard of imprinting, but not the "baby duck factor". (Closest I've come is "What do you get when you put a bunch of ducks in a package? A box of quackers!") Don, thanks so much for polishing up the procedure so nicely. I had noted the CAB function you suggested as an alternative. I will check out your whole script more closely once copied to my system, but -- as always -- you've been most helpful. Note 1123.23, 13-Jun-1995 Pye: If Don's got half a brain, I must have a quarter :-) --------------------------------------------------------- Gosh, I go away from this file for a couple of days, and there's 22 replies! Which reminds me, Digital's Internet gateway is a bit ill at the moment... I don't think I was party to Don's previous requests for this functionality, but it sounds like it would be the sort of philosophy of perfection that we strive for :-) Not to get into a glossy brochure battle, but I think we will be able to offer the same sort of traceability of messages described earlier when we support correlation of read/delivery receipts in V3.2. This will enable you to see what state all the recipients of a message are in (if you see what I mean!) Of course all of this only works if all players are co-operating in the same mail protocol (in our case X.400) and don't see how else anyone can claim this works - unless they have very clever gateways everywhere, and no dumb mail clients. The DECUServe Journal August, 1995 Page 20 Retracting a Mail Message Another approach to the retraction problem is the X.400 superceed functionality which allows you send a message that obsoletes the previous version. This means that you can effectively handle the retraction case more effectively, although with the embarrassing disadvantage that you have to reveal your mistake to all the people that haven't yet read the message! Finally, on the suggestions, we are still adding functionality to the base, and we *are* still interested in suggestions. That's partly why my Manager pays for my time to be in these notesfiles, so I can collect information, as well as try and help you. So suggestions here, or by mail to me, or you can still submit suggestion SPRs, which we *do* get here in Engineering. Note 1123.24, 13-Jun-1995 Bruner: Good plan on DEC's part! -------------------------------- It's good to have you with us again, Graham, and I do gladly accept your word that Digital is still listening and still aiming at improving the base product. The fact that they give you DECUServe access certainly promotes the exchange of important information in BOTH directions! Note 1123.25, 13-Jun-1995 Peach: Comments, fwiw --------------------- I know what you mean Graham -- actually it was a delight to see how this 'note' has grown. The "Baby Duck Factor" -- or in military parlance the BDF is great. And so very true. My problem with doing this for users is being able to do it. By that I mean, can I fairly offer this feature to users if I [or someone else on my team] have to be the ones to do it? Our system hours and access to -- and use of -- is not the same as the staff in-house hours. Do I want to set up a situation where in users can always be changing their minds and recalling messages. I can see messages being recalled but some users had read before the recall was issued -- and then we get calls about why didn't the message get to everyone. I saw user XYZ on the list and yet when I talked to him/her, they said they never got it. Since our training is very limited, I cannot rely on that as a way to handle it. I think this is a great idea but one that will have to be considered for implementation very carefully at each site. In some areas, such as Roger's, it will probably work very well. I don't see it here -- period. But I never pass a good script up! So am downloading it anyway. It The DECUServe Journal August, 1995 Page 21 Retracting a Mail Message might become useful as a 'hidden' option. Ethernet Interconnections -------- ---------------- Abstract: A discussion of connecting a thinwire Ethernet segment to a 10baseT network, and where and how a hub fits into the picture; also includes an interesting book recommendation. Participants: Charlie Byrne, Linwood Ferguson, Michael Mazzoni. Conference: HARDWARE_HELP Note 1895.0, 28-Jun-1995 Byrne" Connecting 10B2 to a 10BT network ---------------------------------------- Our net access here coming out of the router goes 10bt to to a 32 port DEC hub and then many of those ports are used in the local building, a few are run to nearby buildings where they are fanned out with DEC 8 port 10bt hubs. One fellow in one of the adjacent buildings has 4 computers that are already connected to each other with thin wire. Of course now he wants on the net (yes he will have to change his ip addresses of he was using a local ip). He would prefer not to buy 4 new boards as some are for some slightly obscure and possibly obsolete systems that may not even have 10bt available. I guess I could get a 10BT-BNC media converter for $275 from black box. But instead, can I use an 8 port 10BT hub with a BNC connector? ($225) My concerns are: 1) Inmac for examples has 8 port Hub described as follows: User Interface: Eight RJ45 for 10BaseT Backbone Interface: RJ-45 10BaseT Uplink and BNC Seems like I would be trying to use this "backwards", i.e. going from BNC to 10BT. Is that a problem? 2) Do I now have too many segments, this would be the 3rd hub. The DECUServe Journal August, 1995 Page 22 Ethernet Interconnections Note 1895.1, 28-Jun-1995 Ferguson: Transceiver port? --------------------------- I don't know the 32 port DEC hub, but does it have a transceiver port? Generally a lot of hubs are set up for a backbone and either provide thin wire or transceiver ports for that. If so, just run the thinwire over to it and use a $30 or so thinwire transceiver adapter (DECXM from DEC, I think). Note 1895.2, 28-Jun-1995 Byrne: sorry for the dumb question ---------------------------------- Well Duhhh, nevermind. The 8 port hub *has* a BNC port on it. I don't know what I was thinking. Note 1895.3, 28-Jun-1995 Byrne: still confused somewhat ------------------------------ > I don't know the 32 port DEC hub, but does it have a transceiver port? > Generally a lot of hubs are set up for a backbone and either provide > thin wire or transceiver ports for that. So are you saying that you *can* use hubs "backwards" as media converters? In other words, Given an 8 port RJ45 10bt hub, that has BNC and/or AUI "input" ports, You could take that and connect 1 of the 10bt ports to your 10bt net and then the bnc to your 10b2 net? (or via transceiver to any media ethernet)? The documentation for the DEC 90TS hub is no help at all in this regard. Note 1895.4, 28-Jun-1995 Ferguson: Yes ------------- I am not sure there is really a direction involved in this case. Caveat: I am OJT knowledgeable here, and am far from expert, but: Basically the hub is a repeater. It repeats whatever is on the UTP connections to each other and to the AUI or BNC connector. Unless of course the hub is a switch in which case it is a bit smarter but still there is no real direction involved. The DECUServe Journal August, 1995 Page 23 Ethernet Interconnections Consider a different scenario: you have a few dozen hubs in a rack. You connect each to the other via the thinwire BNC connector which serves as (conceptually) a backbone (though of course everything on there is the same speed). Now picture yours overlaying that picture, and all you are really doing is adding another level of repeaters cascading out from one of these repeaters. All you need to do is make sure the various rules for number of repeaters/etc are followed between any two points on the network (and special cases like DELNI's that might be thrown in there). For instance, when connecting a bunch of hubs with thinwire there is never more than two hubs between any two points. With cascading, while there is rarely a need for more levels, people tend to just go buy another and stick it in a closet or something when they need more ports. It ends up being really easy to get another level in there you did not expect (e.g. cascade once in your network room, then have two people cascade locally in an office and you are up to 4 already). Note 1895.5, 6-Jul-1995 Byrne: Ethernet book needed --------------------------- Can someone recommend a good Ethernet book, e.g. "Everything You Need To Know To Configure And Maintain A Mixed Media, Mixed CPU type, Mixed Vendor, Ethernet LAN" Every few years I have to do something and I forget what I knew and have to ask the same old questions. Note 1895.6, 6-Jul-1995 Mazzoni: Here's my favorite --------------------------- "Ethernet - Building a Communications Infrastructure", by Heinz-Gerd Hegering and Alfred Laepple, Addison_Wesley, 1993, ISBN 0-201-62405-2. The best book on Ethernet I've seen. There's probably more detail than you'd ever want to know. Written with the assumption that you already are data communications literate and doesn't waste time starting out with "what is a bit? what is a byte?". Light on a 'rules of thumb approach', heavy emphasis on the underlying standards and engineering issues. The book does not restrict itself to the LAN. Bridges, routers, gateways are extensively treated. FDDI and some MANs are presented as backbones for LAN interconnects. Higher level protocols, PC networks, distributed systems, and network management are also addressed. The DECUServe Journal August, 1995 Page 24 RZ26-VA Transplant RZ26-VA Transplant ------- ---------- Abstract: Trials and tribulations of taking an RZ26 drive that was lifted from a StorageWorks shelf and trying to use it with a disk-starved MicroVAX 3100. Includes a pointer to a DSNlink patch. Participants: Henry Carmichael, Mark Lasoff, David MacLean, Jorma Tapola Conference: HARDWARE_HELP Note 1900.0, 7-Jul-1995 MacLean: trouble transplanting RZ26 out of -VA enclosure -------------------------------------------------------- Having a couple of "surplus" (ie, displaced by larger replacements) RZ26-VA 1-gig disks, and a uVAX3120 with only 300meg, I decided I would remove the RZ26 from the StorageWorks package and plant it into the spare space on the upper shelf of my 3100. I put on the jumpers to address it as device DKB300, plugged it in, turned on the system and find that >>>SHOW DEV recognizes that the RZ26 is at SCSI position B3 (with other disks at A3 and B1, and tape drive at B5), but when I boot up VMS and try to INIT or MOUNT the drive I get DRVERR (fatal drive error) messages. Swapping disks at A3 and B3 made no difference. An obvious possibility is that the drive is actually a dud, but before I go to the trouble of taking another -VA apart, or putting this drive back into -VA so I can test it in the SW500 shelf, ... (1) has anybody had success doing what I am trying to do here? (2) anybody got advice on what I might have neglected in the transplant procedure? Note 1900.1, 7-Jul-1995 Carmichael: If the target is OpenVMS V6.1 ... --------------------------------------------- I had a similar problem with OpenVMS VAX V6.1. The SCSI driver is misinterpreting some information from the drive. If you are on V6.1 (you didn't mention the version), you can obtain a patch from DSNlink. The patch name is VAXSCSI02_061 and has 5 save sets and a cover letter. Hope this will help. The DECUServe Journal August, 1995 Page 25 RZ26-VA Transplant Note 1900.2, 7-Jul-1995 MacLean: I'm using VMS 5.5-2 on the 3100s ----------------------------------------- Sorry, should have mentioned that it is VMS 5.5-2 version on the uVAX3120. Actually I'm running the same 5.5-2 on a 3140 and a 3110 and both use RZ26 disks, but the console firmware may be an older rev on the 3120. I'll check DSIN for SCSI patch that might apply to 5.5-2; thanks for the advice. PS: I put the RZ26-VA back together, and the problem is not with the disk, so it must be related to the 3120 microcode. Note 1900.3, 8-Jul-1995 MacLean: it's the rev level of the drive? ----------------------------------------- I tried the RZ26 in a 3110 that already had one RZ26 installed, and it does not work there either. Problem is obviously the particular rev level of the RZ26, and I will install VAXSCSI patch and see if that cures the problem of getting drive info different from what the driver is expecting. (Aren't standards wonderful? We love them. That's why we have so many different ones!) Note 1900.4, 16-Jul-1995 Tapola: Previously used in VAX? ------------------------------- > Having a couple of "surplus" (ie, displaced by larger replacements) > RZ26-VA 1-gig disks, and a uVAX3120 with only 300meg, I decided I would > remove the RZ26 from the StorageWorks package and plant it into the Was is used in VAX system as RZ26-VA? there might be some difference in jumpers of disk format when used in other systems. You might need to format it. Can you test the drive in BA350 on this particular VAX 3100? Note 1900.5, 17-Jul-1995 MacLean: only ever used the disk on VMS 5.5-2 --------------------------------------------- The drive was used on VAX VMS 5.5-2 cluster, via dual HSJ40 version 2.5, immediately prior to being moved to the 3100. As such, it was set NOTRANSPORTABLE, and appeared as a DU device rather than a DK device on the cluster. I did put it back on the HSJ and SET it TRANSPORTABLE, INITed and used it as a DK, and then tried it again on the 3100; still no go. The only jumper changes of importance are the drive addressing (it The DECUServe Journal August, 1995 Page 26 RZ26-VA Transplant picks it up from cable to backplane in SW enclosure, and I explicitly set it to 3 when external). The only other jumpers related to providing power to the SCSI terminator resistors, which are not even installed on the drive. I don't have any spare SW shelf available, but that is an interesting idea and I will try to borrow one. It would let me check three or four other drives, and confirm that the problem is peculiar to that rev of the RZ26 innards. Unfortunately, all my RZ26-VA drives, with external revs of H and F and at least one other letter, have the same hardware rev on the disk innards. DEC support suggests turning my software call into a hardware call and getting the drive replaced; I'm considering that, or I may just elect to put a non-DEC drive into the 3100 and leave the RZ26-VA box intact as a spare unit. Note 1900.6, 21-Jul-1995 Lasoff: Newer Patch: VAXSCSI03_061 ---------------------------------- > DSNlink. The patch name is VAXSCSI02_061 and has 5 save sets and a ^^^^^^^^^^^^^ FWIW, the latest version is VAXSCSI03_061 ^ |Rev 3. Vanity Domains in URLs ------ ------- -- ---- Abstract: A number of Web service providers will set their customers up so that their home page appears under their own ("vanity") domain name rather than that of the service provider. But how is it that the different names aren't treated as separate entities? Participants: Kevin Angley, Steve Cloutier, Terry Kennedy. Conference: INTERNETWORKING The DECUServe Journal August, 1995 Page 27 Vanity Domains in URLs Note 379.0, 22-Jul-1995 Angley: Vanity domains in URLs - how do they work? -------------------------------------------------- This is either a DNS question, or an httpd question, I'm not sure. Many web pages publishers give you the option of using www.yourcompany.com in the URL, even though you are using their http server (along with many others). OK .. easy enough that the DNS for mycompany.com simply includes a CNAME record for www pointing to the IP address for their http server _... that part I understand. But how is it that you *don't* have to then include a relative path to get to mycompany.com's home page. It would seem that the http server would have only one default home page. Thus, http://www.mycompany.com and http://www.yourcompany.com, if pointing to the same server, would get you the same home page. But it doesn't - it gets you mycompany's or yourcompany's. In exploring this (www.best.com is an example of a provider that does this sort of thing; www.randysoft.com is an example vanity domain pn their server), I find that www.mycompany.com is given a different IP address from www.httpserver.com .. but packets sent to www.mycompany.com do indeed go to www.httpserver.com. So that's a clue, but I'm still not sure how that works. Note 379.1, 23-Jul-1995 Kennedy: --------- > This is either a DNS question, or an httpd question, I'm not sure. It's a Unix sockets programming question 8-) > In exploring this (www.best.com is an example of a provider that does > this sort of thing; www.randysoft.com is an example vanity domain pn > their server), I find that www.mycompany.com is given a different IP > address from www.httpserver.com .. but packets sent to > www.mycompany.com do indeed go to www.httpserver.com. So that's a clue, > but I'm still not sure how that works. If the original HTTP spec said that the host name needed to be passed, all of this would be unnecessary. Normally a program that does a bind() is listening with INADDR_ANY, meaning packets that arrive for that port on any interface will be processed. Also, an "interface" was normally a card and had a single IP address. How- ever, you can create alias addresses for an interface. In that case, you specify an explicit address when doing the bind(). You can then run multiple servers, all listening to their own address, with completely distinct configurations. There are variations on this, including running a single server but getting the address of the local end of the connection and making decisions based on that. The DECUServe Journal August, 1995 Page 28 Vanity Domains in URLs Note 379.2, 24-Jul-1995 Cloutier: How we vector multiple WEB domains to a single system. ---------------------------------------------------------------- As Terry pointed out, this is more of a programming issue... And it's relatively trivial to do, by the way. The upcoming release of the NetNotes / WebNotes server does this. Under some Unix and all Windows-NT operating systems, it is very easy to set up your system to be what they call "multi-homed". The system will respond to more than one network address (and you only need one adapter card to do this). So, now the system is set up to respond to, as an example 199.190.243.200 and 204.87.128.50 Ok, so the addresses are all set up, the DNS entries are in place in the DNS server(s), and the packets will be routed to your system.. Now, you need a Web server which is capable of handling all of these addresses *AND* translating them to some physical directory on your system. I can only speak for the WebNotes server on a particular implementation. The WebNotes server administrator created a simple table of addresses and directories. The server scans the table, and if it finds a match between the destination address and a table entry, it vectors off the request to that directory. From a technical standpoint, there is very little work involved to set this up at least on NT. No fondling of the registry or anything like that.. it's all front-panel user level stuff. The server has to know what to do with it, however, and not all servers have this capability. At first it probably looks like rocket science, but it's really nothing fancy.. just a little creative programming (and a flexible operating system) Article title goes here ------- ----- ---- ---- The DECUServe Journal August, 1995 Page 29 Article title goes here Abstract: Brief abstract of notes stream goes here Participants: J.M. Ivler, James Laferriere Conference: INTERNETWORKING Note 378.0, 21-Jul-1995 Ivler: Learning Internetworking the hard way -------------------------------------------- Now that I'm the Internet weenie I'm learning all sorts of things... and I figured that I would share them with others that are learning the hard way... Now, once you know *why* it happened it's sort of funny, but sometimes the obvious can be hard to live through... Router can see Internet Router can see inner-net [DMZ] netra (firewall) can see Router and Internet netra can see inner-net [DMZ] att (bastion host) can see inner-net [DMZ] att can not see Internet Internet (crl.com) can see Router, and netra Internet can not see att nameservice is through netra, and works for att in resolve. resolve.conf and named look just fine... So, where is the problem? give up? netstat shows the routing table looks just fine... but... when the default was added using the route command it was added as: $ route add host default 100.100.100.1 1 This added the default path as the host default, not the key word default... so a simple fix was... $ route del host default 100.100.100.1 $ route add net default 100.100.100.1 1 This only took 20 manhours to fix... Next weeks task - adding xinetd to replace inetd... look for more war stories... The DECUServe Journal August, 1995 Page 30 Article title goes here Note 378.1, 21-Jul-1995 Laferriere: The advantage to xinetd ? -------------------------------------- Jim, What is the big advantage to xinetd over inetd ? or where can I find this out ? html,doc,faq??? Note 378.2, 22-Jul-1995 Ivler: More on xinetd --------------------- inetd reads the services file, and then either allows or disallows a service. Period. A binary choice. xinetd allows you to allow or disallow a service based on the incoming IP, the time of day, etc. O'Reilly and Associates book on Internet Services seems to be the best reference I have read so far. It, along with a firewall router, provides some levels of security. It also allows us to allow say telnet to come in, but only those machines which are permitted to accept a telnet service will have it active in xinetd. And there will be some that can only be accessed from within the local network (on the inside of the firewall) .v. some that can be accessed from both inside and outside. One key area we will be using it in is in allowing a certain sub-set of the inner machines access to the r* services during working hours. It is just part of the overall picture of site security. Not so much an answer to all of the problems, but it allows a finer granularity of control over the tcp services that are available. Serving CDs with PATHWORKS ------- --- ---- --------- Abstract: Discussion of making InfoServer-mounted CD-ROMs available to PATHWORKS (non-NT) clients, and why that parenthetical exception is there for NT. Participants: Petri Backstrom, Joe Crum, Michael Spatz, James Stang, Elie Uy Conference: PATHWORKS The DECUServe Journal August, 1995 Page 31 Serving CDs with PATHWORKS Note 833.0 Uy: Can Pathwork Serve CD ? --------------------------- Does any one know how to make CD on Info Server be made available for our Pathwork users ? We are currently running Pathwork V5.0 on the VAX, I wonder if there is a way to serve it so that our users that do not have CD on their PC will be able to use it. Note 833.1, 8-Jun-1995 Spatz: Consult DSNlink ---------------------- You have to install an additional Layered Product and issue a mount command with a qualifier. I do not recall the exact product. I believe that there is an article on DSNLINK Note 833.2, 10-Jun-1995 Backstrom: ----------- PW clients can also connect directly to InfoServers; just select "Virtual Disk Services" in PWSETUP, go into "Detail" and select CD-ROM services. All you need to connect to InfoServers from VMS is included with V6, but if you use V5.5-2 then you need the F11CD kit from the InfoServer directory on the CONDIST CD's. Note 833.3, 14-Jun-1995 Crum: Not under NT, apparently ------------------------------ PW clients under DOS/Win3.1 (and maybe Win95), that is. PW for for Windows NT (my client OS) does *not* include LAD, or so I have been told by Digital. If this is not true, please correct me. Note 833.4, 15-Jun-1995 Uy: It works.. Thanks --------------------- I tried PWSETUP and I was able to use the CD on InforServer. Thank you very much for all your help.. The DECUServe Journal August, 1995 Page 32 Serving CDs with PATHWORKS Note 833.5, 16-Jun-1995 Stang: Why not copy to magnetic media? -------------------------------------- This may seem odd but we have basically taken a different approach to mounting CD's to be used on the network (well by Pathworks users anyway). A Segate 9GB drive is capable of storing about 15 CDs worth of information on it and only uses up one SCSI address. We are currently using a 2.1 GB Barracuda to serve up 2 databases. The software simply looks like a 'share' and most PC software doesn't really need MSCDEX to work. As always, your mileage may vary but if you have 600Mb to 1.2 Gb free storage on a VAXEN try it out. (BTW a Baraccuda makes even a 6x CDRom drive seem positively snail like ;-) Note 833.7, 27-Jun-1995 Backstrom: ----------- Re: .3 Right. Not under NT; there are no LAD/LAST protocol implementations for NT. PATHWORKS-to-PATHWORKS Copy ---------------------- ---- Abstract: Consideration of ways to move documents from VAX-based PATHWORKS to an NT server, also running PATHWORKS. Two variations on the theme of FAL16 offered as a solution. Participants: Ta Fuh Chiam, Mike Gozaloff, Michael Spatz. Conference: PATHWORKS Note 840.0, 7-Jul-1995 Gozaloff: Can VAX PW server access WNT PW server? ------------------------------------------------- We need to put together a program that moves some ALL-IN-1 documents from a VAX Pathworks server (probably 4.1 or 4.2) to a Pathworks NT server. The directory structure on the NT shares should mimic the The DECUServe Journal August, 1995 Page 33 PATHWORKS-to-PATHWORKS Copy folder structure in ALL-IN-1. For extra credit, the customer would like the NT shares to be automatically created and access set. The initial approach would be to write some VAX ALL-IN-1 code to get the file and directory names, create the shares (if possible) and copy the files. Is there a way the VAX Pathworks server can write the files on the NT server? Can it also create the shares and modify the access rights? If the VAX approach proves too difficult, we can use TeamLinks on a PC to move the documents from the file cabinet to a mapped drive on the NT system. Can a PC program create shares on the NT server and modify the access? Note 840.1, 8-Jul-1995 Chiam: ------- A question, I am curious, how are you going to access these ALL-IN-1 documents from the NT server? Note 840.2, 10-Jul-1995 Gozaloff: Just simple file sharing ---------------------------------- No magic, they're setting up file services using NT Pathworks server and Windows (or WfW) clients. When the files are exported from ALL-IN-1 they will be limited to short file and directory names but most of the document titles are already short. Originally the files came from an 8.3 file system and were imported into ALL-IN-1. Most of the files are template word processing documents and are set read only. Note 840.3, 11-Jul-1995 Spatz: VAX to NT copy --------------------- I believe this is what you are looking for, but in no way does this fit what our organization is looking for. We would like to be able to copy files to an NT server utilizing the DECnet protocol, this would require that there be a process running on the NT BOX. Currently this requires an interactive login and FAL16 running at a command prompt. To set this up perform the following: ncp ncp> define obj fal file %P/fal%U c:> fal16 Then copy files from a vax to the NT. The DECUServe Journal August, 1995 Page 34 PATHWORKS-to-PATHWORKS Copy Note 840.4, 12-Jul-1995 Spatz: fal under spawner ------------------------ A better and more realistic approach will let the native FAL start automatically under the spawner for each request. define the obj fal ncp > define obj fal number 17 file "%P/fal %U" This works InfoServer CDs to Windows NT ---------- --- -- ------- -- Abstract: As mentioned above, LAST is not available for Windows NT, so no direct connections to InfoServer CDs are possible. Here, though, is a way around that restriction, using an OpenVMS system as intermediary. Participants: Petri Backstrom, Mitchell Nussbaum, Dan Singleton, Alan Striegel. Conference: PATHWORKS Note 829.0, 26-May-1995 Nussbaum: Can Windows NT connect to an InfoServer? -------------------------------------------------- I'm thinking of installing Windows NT client on my PC (with Pathworks 4.1 client), but I before I do this, I need to be sure I'll have some way to connect to CD services and hard disk partitions on our InfoServer. I've looked in every piece of documentation that I can get my hands on, but I can't find any references to LAD or LAST or, for that matter, to InfoServers. Does this mean that Digital does not support InfoServer access from Windows NT clients? Or is there some strategy that's been eluding me? Note 829.1, 28-May-1995 Backstrom: ----------- There is no LAD/LAST for Windows NT. In other words, Windows NT systems cannot connect directly to InfoServers (but a VMS system The DECUServe Journal August, 1995 Page 35 InfoServer CDs to Windows NT that can could be offering the InfoServer devices as file services). Note 829.2, 5-Jun-1995 Striegel: Serve InfoServer CDs as Share from VMS? How? ------------------------------------------------------- Can someone summarize quickly the actual steps (VMS commands) to make this work? Note 829.4, 4-Jul-1995 Singleton: I believe it is not possible currently ------------------------------------------------- I have looked for this and I have not been able to find anything! Bottom line (I believe) is that you can not use NT and connect to an Note 829.5, 4-Jul-1995 Backstrom: ----------- InfoServer> CREATE SERVICE service ... $ MC ESS$LADCP BIND service $ MOUNT... Edit LANMAN.INI and add autoshare=device=shortname. Start PW server. Details in the appropriate manuals (InfoServer, VMS, PW). Intel PC Booting ----- -- ------- Abstract: A discussion of the rules used by (most) Intel PCs to determine which device should be the boot drive. Includes yet another valiant attempt to understand Intel things in Macintosh terms and vice versa -- two different worlds. Participants: Dale Coy, Terry Kennedy, Larry Kilgallen, Jeff Killeen, Mike Taylor. The DECUServe Journal August, 1995 Page 36 Intel PC Booting Conference: PERSONAL_COMPUTING Note 722.0, 10-Jul-1995 Kilgallen: How does an Intel machine know from where to boot ? -------------------------------------------------------------- On a Macintosh one selects what is essentially a SCSI unit number using a control panel application and that controls what hard disk the machine will try first when it boots. How is the selection between various hard disks handled on an Intel machine, given the variety of hard disk connection mechanisms - some IDE and some SCSI ? Is the mechanism consistent across all manufacturers, ages of machines, etc ? Note 722.1, 10-Jul-1995 Kennedy: The rules ------------------ Under MS-DOS, drives can be subdivided into "partitions" which have their own drive "name" (letter, like D:). The order is (first to last): Primary partition on drives supported by the motherboard BIOS Primary partition on drives supported by controller BIOS in ascending ROM address order Additional partition(s) on drives supported by the motherboard BIOS Additional partition(s) on drives supported by the controller BIOS in ascending ROM address order Additional drives supported by DOS device drivers (unless drive letter overridden in config) Network drives (unless drive letter overridden in network config) The motherboard BIOS supports ST-506 interface drives (MFM, RLL, and IDE) while the controller BIOS is used for ESDI, SCSI, and some RLL drives. So, some examples (info in parenthesis indicates physical drive number and partition order - not to be confused with drive letter under DOS): IDE drive alone, only 1 partition: C: IDE drive w/ SCSI drive, 1 partition on each: IDE C:, SCSI D: 1 IDE, 1 SCSI, 2 partitions on each: IDE(0a) C:, SCSI(0a) D:, IDE(0b) E:, SCSI(0b) F: A really hairy scheme - 2 IDE drives, 3 partitions on each, 1 SCSI drive w/ ROM at C800, 2 partitions, 1 ESDI drive w/ ROM at DC00, 3 partitions, one CD-ROM drive: IDE(0a) C:, IDE(1a) D:, SCSI(0a) E:, ESDI(0a) F:, IDE(0b) G:, IDE(0c) H:, IDE(1b) I:, IDE(1c) J:, SCSI(0b) K:, ESDI(0b) L:, ESDI(0c) The DECUServe Journal August, 1995 Page 37 Intel PC Booting M:, CD-ROM N: > Is the mechanism consistent across all manufacturers, ages of machines, > etc ? Yes, with 3 exceptions: Future Domain brand SCSI controllers (and maybe some other controllers) scan down from ID 7 to 0 instead of up from 0 to 7. This means that if you have SCSI ID's 0 and 1 for drives, 1 is C: and 0 is D:. This is considered a bug by many people. Some motherboard BIOS implementations scan for adapter BIOS ROMs from the top down instead of the bottom up. This would affect cases with more than one controller BIOS (unusual). Some motherboard implementations on "bounded" PC's may include motherboard BIOS support for ESDI/SCSI drives. If such a drive is controlled by the mother- board BIOS instead of a controller BIOS, it's treated as an IDE drive for sequence purposes. Can you tell I get asked this a lot? 8-) Note 722.2, 11-Jul-1995 Coy: 32-bit file access query ----------------------------- Since this looks like a good place for detail: > Some motherboard implementations on "bounded" PC's may include motherboard >BIOS support for ESDI/SCSI drives. If such a drive is controlled by the mother- >board BIOS instead of a controller BIOS, it's treated as an IDE drive for >sequence purposes. I presume I have one of these (AMI BIOS, etc.). I have no IDE drives -- just an Adaptec SCSI Controller (1542) with two SCSI disk drives. The first drive (0) is divided into 2 partitions (512 and "the rest"), that show up as C and D (and BIOS control). The second drive (1) is divided into 3 partitions that show up as E,F,G. My question is about WFWG 3.11 32-bit file access. When I enable 32-bit file access, _only_ the D drive shows up as using it. I understand why E,F,G do not (controller BIOS), but I can't understand why *C* would not use 32-bit file access, although *D* will -- and they are both on the same physical drive, and the same BIOS. The DECUServe Journal August, 1995 Page 38 Intel PC Booting Note 722.3, 11-Jul-1995 Taylor: What does fdisk do? --------------------------- Is the order in .1 the default order? If you use fdisk to mark one of the partitions as active, is that overriding the order? Or, is it doing something else? This may be of some help to .0: fdisk is a DOS utility that can be used to view partitions and select an active partition. For PC troubleshooting and maintenance it is usually a good idea to have a bootable DOS floppy with fdisk. (I think KGB talked about that in DEC Professional one time, saying he kept several floppies in various places.) I am familiar with two partitions on one IDE drive - one FAT and one NTFS. When DOS is running, I can use fdisk to mark the NTFS partition as active so NT boots at the next boot. When NT is running, I can use Disk Administrator to mark the FAT partition as active so DOS will boot the next time. How does fdisk fit in with the rules? Note 722.4, 11-Jul-1995 Coy: C = 512 = active --------------------- > Is the order in .1 the default order? If you use fdisk to mark one of > the partitions as active, is that overriding the order? Or, is it > doing something else? If you are referring to my query in .2... That is the default order (C=512, D=remainder) and the C-partition is marked active (shows that way in fdisk). That's the way I thought it should be. Note 722.5, 11-Jul-1995 Kilgallen: But how does one _select_? ------------------------------------- I guess I don't understand, but that looks like a _default_ order in which the console code would search the drives for an operating system. What I was wondering is how one "selects" which drive the console code (BIOS ?) will boot from. So far as I can see this should be independent of operating system disk structures, since one could have Linux on one drive and Windows 95 on another. On the Macintosh, the little program one runs writes to EEPROM which is used by the machine on the next boot to make a selection. The DECUServe Journal August, 1995 Page 39 Intel PC Booting Note 722.6, 12-Jul-1995 Killeen: --------- > How is the selection between various hard disks handled on an Intel > machine, given the variety of hard disk connection mechanisms - some > IDE and some SCSI ? > > Is the mechanism consistent across all manufacturers, ages of machines, > etc ? NT puts a boot loader on the first drive. That boot loader will let boot whatever wherever. So the technology exists... Note 722.7, 12-Jul-1995 Kennedy: Oh ----------- > I guess I don't understand, but that looks like a _default_ order in > which the console code would search the drives for an operating system. Ahh. I misinterpreted your question as "what are the drive assignments after booting". > What I was wondering is how one "selects" which drive the console code > (BIOS ?) will boot from. So far as I can see this should be independent > of operating system disk structures, since one could have Linux on one > drive and Windows 95 on another. It boots from the first drive that exists in the order I gave a few notes ago. Note that that's the "primary bootstrap", which in vanilla MS-DOS looks for the first partition on that disk that has the "bootable" flag set and then loads that particular bootstrap. Alternative primary boots exist which will do a number of things, including asking you which partition to boot (so you're not hardwired to the one that's marked bootable), which drive to boot from, etc. Under DOS there are some restrictions - there is no way (a la VMB) to pass boot info to the next stage, so it's possible to configure a primary bootstrap to boot some alternate disk/partition and then have the next phase of the boot die with a "how did I get here?" sort of error because a lot of code assumes that if it got booted, it has to be on the first bootable disk. Note 722.8, 12-Jul-1995 Kilgallen: Ah, now we're getting into what I needed ! ----------------------------------------------------- > Under DOS there are some restrictions - there is no way (a la VMB) to pass > boot info to the next stage, so it's possible to configure a primary bootstrap > to boot some alternate disk/partition and then have the next phase of the boot > die with a "how did I get here?" sort of error because a lot of code assumes > that if it got booted, it has to be on the first bootable disk. The DECUServe Journal August, 1995 Page 40 Intel PC Booting Is that "possible" or "certain" ? If a real machine were configured with Jeff's suggestion (Windows NT first), would any DOS disk to which the boot was redirected fail? Are there some DOS disks which would not fail? What determines success or failure? Does OS/2 do any better than the DOS example of "how did I get here, crash"? Any thoughts on SCO/Unix or Linux regarding that particular confusion ? DCL Error Trapping --- ----- -------- Abstract: In the following, a simple question gets answered quickly and accurately, which is nothing new on DECUServe, and then followed up by the original poster with a detailed explanation to help save other folks from similar troubles -- also nothing new on DECUServe. The main topic is the ON ERROR handling in a DCL routine. Participants: Larry Clegg, Dale Coy, Kevin Gilmore, Chuck McMichael, Larry Rosenbaum. Conference: VMS Note 2549.0, 7-Jul-1995 Gilmore: Simple (?) DCL question. --------------------------------- I have a question re: using $SET ON (or $SET NOON). Is there an associated command that could tell me the characteristics of the ON condition handler? Something like $SHOW STATUS ON ? If the answer to this is RTFM, please supply a page number ;^) Note 2549.1, 7-Jul-1995 Clegg: Lexicals! ---------------- Try HELP LEXICAL F$ENVIRONMENT ITEM - some of the information you want is in there. Enjoy. The DECUServe Journal August, 1995 Page 41 DCL Error Trapping Note 2549.2, 7-Jul-1995 McMichael: ----------- I haven't used it, but try F$ENVIRONMENT("ON_SEVERITY") Note 2549.3, 7-Jul-1995 Gilmore: Guess I misunderstood myself... ---------------------------------------- Re: .1 & .2. Thank you for the quick response. I can think of a few service organizations the could take a cue from DECUServe members quick and accurate responses. I tested a few things and found to my surprise that I had been using the ON command in a manner that won't work. So in the spirit of conciseness, here's the RDV (Readers Digest Version): The ON command can only track _one_ condition, the most recently defined one. The $SEVERITY symbol takes on the status of the associated command, not the condition just detected, and the F$ENVIR("ON_SEVERITY")/ON condition get reset to their default values (a gotcha). For anyone that wants some more explanation and an example, read on... (but beware, I've been accused of being long winded) Okay, misconception clearup! I have been fooling myself about the use of the $ON command and thought some other people would appreciate my learning experience. Some of you know this and understand this already, but I'm sure there are some others who don't (or thought we did). __Misconception: By executing the ON command for each of the three possible conditions you can have different responses for different conditions. (ex: $ ON SEVERE THEN GOTO severe_condition $ ON ERROR THEN GOTO error_condition $ ON WARNING THEN GOTO warning_condition ) *Reality: Use of the ON command supersedes the previous definition (or default). In the above example if you tried to open a file that didn't exist, the command executed would be "GOTO warning_condition" even though attempting to open a file that doesn't exist is an error (%DCL-E-OPENIN, see example below). The ON warning superseded the previous ON error command. And, by using the ON command you are specifying the _minimum_ condition necessary to execute the associated command. ---------------------------------------------------------------------------- __side-bar misconception: You can use the $SEVERITY symbol to check for the condition after it has been detected. (lets say you did a WRITE SYS$OUTPUT instead of a GOTO) The DECUServe Journal August, 1995 Page 42 DCL Error Trapping *that's false: When the condition is detected an associated command is executed, thereby resetting the $SEVERITY symbol. The symbol only tracks the last executed command. Of course that's standard behavior, but it may not be very obvious when you're trying to solve a problem. (and it's the only DCL command I know of that lets you execute another command on the same line, other than IF-THEN) ----------------------------------------------------------------------------- #Example: This example shows (at least) two things. The first is the resetting of the ON condition and the resetting of the $SEVERITY symbol (and item code). $ on error then $write sys$output "error" $ on warning then $write sys$output "warning" $ what = f$envir("on_severity") $ sh sym what WHAT = "WARNING" $ sh sym $* $RESTART == "FALSE" $SEVERITY == "1" $STATUS == "%X00030001" $ open/read file1 nothing.there %DCL-E-OPENIN, error opening USR_SCRATCH:[GILMORE_K]NOTHING.THERE; as input -RMS-E-FNF, file not found warning $ sh sym $* $RESTART == "FALSE" $SEVERITY == "1" $STATUS == "%X00010001" $ what = f$envir("on_severity") $ sh sym what WHAT = "ERROR" Well, I hope that helps save someone from a future gotcha. Note 2549.4, 7-Jul-1995 Coy: You're welcome ------------------- > Re: .1 & .2. Thank you for the quick response. I can think of a few service > organizations the could take a cue from DECUServe members quick and accurate > responses. 21 minutes for two correct responses, and NO incorrect responses, is fairly slow for DECUServe. :-) Your thanks are appreciated, and thank YOU for supplying the complete digest version. It will be a lot of help to others, in the future. > The ON command can only track _one_ condition, the most recently defined one. Don't feel bad about misunderstanding that. It's not exactly The DECUServe Journal August, 1995 Page 43 DCL Error Trapping intuitively obvious, and most of us, myself included, have made the same mistake. Note 2549.5, 10-Jul-1995 Rosenbaum: ON condition THEN GOTO label preserves $SEVERITY ----------------------------------------------------------- If the conditional command is a GOTO then $STATUS and $SEVERITY will retain the values from the command that failed: $ ON ERROR THEN GOTO ERROR $ some command that fails _... $ERROR: $ SHOW SYMBOL $SEVERITY Low Priority Batch Jobs --- -------- ----- ---- Abstract: Various aspects of tuning a system to minimize impact of batch jobs on response time. The catch? The jobs are I/O-intensive. But how many DECUServe regulars can resist a challenge? Participants: David Campen, Rick Carter, Dale Coy, Arnold De Larisch, Linwood Ferguson, Jamie Hanrahan, Bob Hassinger, Larry Kilgallen, Patrick Stair, Jorma Tapola, Chris Wesling. Conference: VMS Note 2553.0, 14-Jul-1995 Carter: How to make low-priority batch not be "felt" by others? --------------------------------------------------------------- A colleague across campus is having a problem with a batch queue. The setup: VAX 4000, VMS 6.1. Three queues: SLOW - priority 1 MEDIUM - priority 2 FAST - priority 3. Someone submits a big job on queue SLOW, and everything else on the system comes to a standstill and it takes 90-some percent of the CPU. A theory we've been playing with is that memory swapping is still doing The DECUServe Journal August, 1995 Page 44 Low Priority Batch Jobs it. Which brings me to the general question: What other things should one set aside from priority to ensure that a low-priority queue really acts as one would think, and only takes resources not used by anyone else? Note 2553.1, 15-Jul-1995 Ferguson: Hard to set queues to prioritize I/O ---------------------------------------------- There are things you simply cannot set (at least for sure). A low priority job that is saturating a disk with I/O is going to interfere with a high priority job that needs the same disk. While you can play with some SYSGEN parameter that penalizes a job for I/O, this really is of negligible help. Memory contention is controllable, though it seems unlikely to be your problem unless the system as a whole is memory poor. That is not a prioritization issue but a tuning issue. Another aspect is priority aging. I forget the exact algorithm, but basically a low priority job waiting for certain events (like I/O) has its priority increased so that when it occurs, it will get the CPU quickly. Using 1-3 for a range of priorities is somewhat of a subtle boost for a job that is not strongly compute bound. We use 1 for batch, 3 for normal interactive, 5 for "pretty urgent" and 10 for processes where near real time response is required. This puts the 10 out of the range that the 3's ever age up to. Note that the priority is dropped again when it gets the CPU -- this is not a huge issue, but is noticeable in some circumstances. Point out to your users (as I have never been able to convince mine) that when there is ONLY ONE JOB running it really does no good to keep upping the priority. :-( Note 2553.2, 15-Jul-1995 Kilgallen: Interference can be minimized by using dumb disks ------------------------------------------------------------ > There are things you simply cannot set (at least for sure). A low > priority job that is saturating a disk with I/O is going to interfere > with a high priority job that needs the same disk. To a certain degree, this depends on the disk. VMS can order IRPs on the queue according to process priority, but if you have a disk with tagged command queueing (or a "smart" Hxx controller doing the same thing) ordering of requests which have been passed off that far is going to depend only on seek and rotational optimization rather than on the non-hardware concept of process priority. The DECUServe Journal August, 1995 Page 45 Low Priority Batch Jobs Basically, priority matters until the request gets sent out from the CPU, so if you have a disk system which holds multiple outstanding requests outside the CPU you lose the effect of priority. Note 2553.3, 15-Jul-1995 Ferguson: Really? Do I have to ask it to? ------------------------------------------ I had never heard that. Does "can order" mean it does order them that way? Or is this something you turn on and off. Note 2553.4, 16-Jul-1995 Kilgallen: If memory serves, you cannot prevent it -------------------------------------------------- I don't have listings handy, but check out EXE$QIODRVPKT which FDT routines invoke to queue the IRP. As I recall, that routine inserts the IRPs on the queue in priority order, and the only way a driver writer could avoid that would be to write their own replacement function. It has been a _long_ time since I looked at this. I was working on a driver just last week, but it was for a non-shared device so the influence of process priority on queue ordering was just not an issue. Note 2553.5, 16-Jul-1995 Tapola: Working set for queues can be altered --------------------------------------------- > Memory contention is controllable, though it seems unlikely to be your > problem unless the system as a whole is memory poor. That is not a > prioritization issue but a tuning issue. I agree that memory usage should be seen as tuning issue, but... If you want to play with memory given to jobs in certain queue, there are "set que"-parameters for working set. Be ware thought that by reducing working set for process that needs larger working set, you probably end up with lots of unnecessary overhead. Generally I would recommend working set parameter for queues only to be used for increasing working sets, not reducing them. Note 2553.6, 17-Jul-1995 Hanrahan: Oh yes it can... -------------------------- Re queueing of IRPs... EXE$QIODRVPKT's priority ordering of IRPs applies only to drivers that use the standard UCB busy bit and wait queue. DUDRIVER does not do that; its start IO routine starts out by clearing the busy bit! So it bypasses normal IRP queueing even though it uses The DECUServe Journal August, 1995 Page 46 Low Priority Batch Jobs QIODRVPKT. Cute, eh? More: DUDRIVER does not appear to refer to the priority fields in the IRP or CDRP. It's possible that the port drivers such as PADRIVER, or the underlying hardware, might order CDRPs by requesting process priority. Note 2553.7, 17-Jul-1995 Kilgallen: ----------- > More: DUDRIVER does not appear to refer to the priority fields in the IRP or > CDRP. It's possible that the port drivers such as PADRIVER, or the underlying > hardware, might order CDRPs by requesting process priority. That would make sense, since that is the point where there is knowledge of how many requests the far end can handle at a time (perhaps not much of a problem if your disk controller is the MSCP server in another system, but quite relevant if the far end is a humble RZ26 doing tagged command queueing). I think we can presume the RZ26 does not arbitrate based on process priority, but it would be a reasonable thing for the MSCP server to do (not that I claim it is done). Note 2553.8, 17-Jul-1995 Campen: I would think priority boost is doing it. ------------------------------------------------- > Someone submits a big job on queue SLOW, and everything else on the > system comes to a standstill and it takes 90-some percent of the CPU. > A theory we've been playing with is that memory swapping is still doing > it. When you say "everything else on the system ", I presume you mean everything in the other two batch queues. I wouldn't think that paging is a problem as I would expect that the priority 1 job is doing lots of I/O and getting priority boosts while the priority 2 and 3 jobs are doing no I/O and getting no priority boosts. If you really want three distinct levels of batch priority you need to separate them by at least 3, possibly 4 priority units - e.g. 1, 4, 7 or 1, 5, 9 and then if you have any interactive users you will have to also raise their priority. Note 2553.9, 17-Jul-1995 Carter: I'll have him play with interactive priority ---------------------------------------------------- > When you say "everything else on the system ", I presume you mean > everything in the other two batch queues. I wouldn't think that paging Actually, he says interactive users really get a slowdown too. That's priority 4. Maybe I'll have him try to set interactive to 6 for a user or two and test (I'll make sure he knows to test it out and what the ramifications might be); AND that he'll have to set it up to 6 for The DECUServe Journal August, 1995 Page 47 Low Priority Batch Jobs everyone if it works). I don't think they care so much about the other batch jobs, at least not now, but I'll explain the principle. Note 2553.10, 17-Jul-1995 Wesling: Lower queue instead of raising users --------------------------------------------- > Actually, he says interactive users really get a slowdown too. That's > priority 4. Maybe I'll have him try to set interactive to 6 for a user > or two and test (I'll make sure he knows to test it out and what the > ramifications might be); AND that he'll have to set it up to 6 for > everyone if it works). Rather than set users up to 6, why not set queue SLOW down to 0? That should ensure that both your prio 3 batch jobs and prio 4 users can pre- empt the CPU from the job in SLOW at any time. Note 2553.11, 17-Jul-1995 De Larisch: Adjust Quantum and Batch Job Priority ------------------------------------------------- > Rather than set users up to 6, why not set queue SLOW down to 0? That > should ensure that both your prio 3 batch jobs and prio 4 users can pre- > empt the CPU from the job in SLOW at any time. I fully agree ... this normally does the trick for us (particularly for CPU intensive jobs) on Alpha systems. You may wish to consider lowering the quantum value from the default. An awful lot of instructions can be executed (on an Alpha system) in 200 Ms (the default Quantum). For interactive users its really annoying to have to 'wait' for the CPU ... and since it IS a dynamic parameter, you can adjust it 'live'. Try cutting it in half (to 10 units of 10Ms) along with placing the slow batch jobs down to priority zero. Note 2553.12, 21-Jul-1995 Hassinger: Memory? ------------------ It is too far back now for me to find, but I am pretty sure there where discussions here on DECUServe back in the early days about memory contention problems in cases more or less like this. A low priority job that required a large amount of memory could could cause serious performance problems for interactive priority level users. That was years ago, when I was managing an 8 Mb 750 that had this sort of problem not to mention a lot of other performance problems based on memory shortages). Since I have not seen discussions of the subject for a long time I suspect the problems may have been resolved or worked around - or maybe everyone has enough memory now to avoid the The DECUServe Journal August, 1995 Page 48 Low Priority Batch Jobs situation... Note 2553.13, 24-Jul-1995 Carter: $$ More $$ Memory $$ is what they plan to do ---------------------------------------------------- > a long time I suspect the problems may have been resolved or worked > around - or maybe everyone has enough memory now to avoid the > situation... Actually, they are in the process of buying more memory. I'll try to find your old notes -- sometimes difficult with FW (search for "VMS" and "Memory" might give a few more hits than I'd like :-)) Note 2553.14, 25-Jul-1995 Stair: I like priority zero --------------------------- I've usually had good response setting a batch queue or an individual process down to zero priority. While your jobs may be more resource intensive, I've seen good results (not _perfect_, but good) on both I/O- and cpu-intensive jobs. Note 2553.15, 25-Jul-1995 Coy: OK now ----------- That's OK on recent versions of VMS. Not generally a good idea on really-old versions (it would time-slice with the Null process) Note 2553.16, 25-Jul-1995 Hassinger: ----------- > I'll try to find your old notes -- sometimes difficult with FW (search > for "VMS" and "Memory" might give a few more hits than I'd like :-)) It might be that what I was recall was actually on the Pageswapper system. Since I have a feeling Larry K (proprietor of that system for a long time - maybe still, I have not tried in a long time...) may be able to help. I suspect he was involved in the discussion no matter which system it was on. The DECUServe Journal August, 1995 Page 49 Low Priority Batch Jobs Note 2553.17, 25-Jul-1995 Kilgallen: Pageswapper Notes ---------------------------- At Dale Coy's suggestion, I am working (slowly) at moving a copy of the Pageswapper notes to DECUServe for read-only access. WWW Advertising --- ----------- Abstract: Begin with a question about a school newspaper wanting to publish a WWW edition and wondering how to handle advertising. Add several interested and knowledgeable DECUServe subscribers. The result is the following notes stream, with several useful points to ponder. Participants: Charlie Byrne, Pierre Hahn, Bob Hassinger, J.M. Ivler, Ed Kozam, Chester S'groi, Don Vickers. Conference: WWW Note 46.0, 23-Jul-1995 S'groi: HTTPD: Advertising on WWW newspaper editions ---------------------------------------------------- Our university recently got a webserver running DEC UNIX. Our local school newspaper wants to go online. We have looked around and noticed that other online papers do *NOT* have advertisements even though they are otherwise complete as far as articles and photos are concerned. The question is: Are there agreed upon conventions that exist for newspapers that want to publish on the WWW and is there something in these regulations that prevent a school newspaper from inserting ads that otherwise appear in its hardcopy version? Specifically, are they prohibited from placing advertisements on their online edition when their online edition is on a WWW server maintained by a university or other educational institution? Our local admins are still in the formative stage as far as this issue is concerned but we don't want to get our site in trouble by violating some unspoken Internet rule. Note 46.1, 23-Jul-1995 Hahn: I would not ----------------- Most University papers have advertisement/coupons for local concerns, they would hold no interest (in fact they would be The DECUServe Journal August, 1995 Page 50 WWW Advertising a bother) to a reader from outside the area. Would net printed coupons be redeemable? As a philosophical discussion - what would you charge your advertisers - you now have a readership/distribution which is much enlarged! This looks like a ball of snakes Note 46.2, 23-Jul-1995 Kozam: The new net etiquette ---------------------------- The whole issue of commercialism on the net is becoming a moot point since NSF is not paying the bill. There are many examples of advertising on the internet, so I don't think you are violating any rules or any part of the newly emerging net etiquette. Note 46.3, 23-Jul-1995 Ivler: Market the net... not advertise -------------------------------------- Rethink "advertising". Who is the audience? Can the paper have someone associated to it who would like the "marketing" link? For instance, if the paper is for "students", then the advertising should be a link to the web-page for the bookstore. If the WWW paper is for alumni, then you might want to consider a "sponsorship" where the alumni association sponsors the page, or some company that makes alumni merchandise does... First: Determine the market Second: determine if the audience [yes, these two can be *very* different] Third: think "marketing" not "advertising", and determine if there is a market. Finally, base costs for those interested on the "hits" on the page that they "sponsor" and then separately on their page. That way they pay for their actual message delivery... Note 46.4, 23-Jul-1995 Hassinger: More than one class of net access being sold? -------------------------------------------------------- > The whole issue of commercialism on the net is becoming a moot > point since NSF is not paying the bill. There are many examples of > advertising on the internet, so I don't think you are violating any > rules or any part of the newly emerging net etiquette. The DECUServe Journal August, 1995 Page 51 WWW Advertising I have not dealt with this myself, but in my company we have been told there are two kinds of access to the net available. More or less they are "research" and "commercial", with the first priced lower than the second. I don't know if that is a general situation, or something specific to our companies provider, or even if it is an accurate representation of the current situation. However, it could suggest a need to check what the terms of your Internet connection are before going into it. There are many people on the net who have not yet accepted the new commercialism and seem to wait for chances to jump all over anyone who looks like they even _might_ be doing something outside the rules. Note 46.5, 23-Jul-1995 Vickers: Makes ads appeal to the readers and optional, too ---------------------------------------------------------- A significant portion of the pages on the WWW are advertising of some sort or another. I agree completely with the points made by Mr. Ivler in .3. Placing print style ads in the 'paper' would not be wise. Changing them to suit the electronic hypertext environment can actually enhance the value of the advertisements if they appeal to the market and audience. I have noticed a few magazines and news type sites that do carry advertisements. Most of these offer links to the ads so that the reader makes a choice to read the ads rather than being forced to 'endure' them to get to the content. Note 46.6, 24-Jul-1995 Byrne: Qualifying Visits ------------------------ With increasing commercialism on the net, the definition of a "hit" is taking on increased importance to those involved. This is why many newer pages are now requiring you to "register" before you get in to look, even if it is free. These visits can be counted as "good" hits, not generated by some index searching program or such. Note 46.7, 24-Jul-1995 Ivler: Demograhics != hits -------------------------- Registration is also used to see more about the contact. For instance, I use iAudit on the Magic Castle Homepage. This counts hits (and also reloads) by the pulldown on an image. The great part is that it avoids "dup hits" as the image is most likely cached so returning to the page doesn't generate another hit. This doesn't tell me much about the visitor... on the other hand, I can ask a few questions and get visitor demographics that add to The DECUServe Journal August, 1995 Page 52 WWW Advertising the value of the individual numeric statistics. That is what registration offers, that straight numbers don't. Demographics. How you capture statistics is as important as the statistics that you capture. I, by the way, recommend iAudit (currently a free service) as a way to do access counting and statistics generation if you don't have access to the log files. PS: Hits != demographics Demographics != reality These Demographics are generated by people that *want* to participate. One thing I like to look at when people do Demographics Collection, it is interesting to see (and it is possible to see) who came, got the page and left .vs. those that responded. 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. The DECUServe Journal August, 1995 Page 53 WWW Advertising 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 griffith@ac.grin.edu +1 515 269 4901 (normal office hours, U.S. Central time) +1 515 269 4936 FAX