From:	ADVAX::"mcnc!VM1.NoDak.EDU!toon%NEWS.SARA.NL" 27-APR-1991 13:01:20.21
To:	Multiple recipients of list ANU-NEWS <ANU-NEWS@NDSUVM1>
CC:	
Subj:	NEWGRPFILE/NEWITMFILE can create non-fragmented NEWS.GROUPS/NEWS.ITEMS

Received: by ADVAX.DECnet (utk-mail11 v1.5) ; Sat, 27 Apr 91 13:00:47 EDT
Received:  from mcnc by ge-dab.GE.COM (5.61/GE-DAB 1.15) with UUCP
	id AA08621 for ; Sat, 27 Apr 91 12:27:39 -0400
Received: from VM1.NoDak.EDU by mcnc.mcnc.org (5.59/MCNC/3-21-91)
	id AA08710; Thu, 25 Apr 91 20:08:09 -0400
	for ARISIA.DNET.ge.com!EVERHART
Message-Id: <9104260008.AA08710@mcnc.mcnc.org>
Received: from NDSUVM1.BITNET by VM1.NoDak.EDU (IBM VM SMTP R1.2.1MX) with BSMTP id 4180; Thu, 25 Apr 91 19:07:51 CDT
Received: from NDSUVM1.BITNET by NDSUVM1.BITNET (Mailer R2.07) with BSMTP id
 7587; Thu, 25 Apr 91 19:07:50 CDT
Date:         Thu, 25 Apr 91 14:59:02 GMT
Reply-To: mcnc!VM1.NoDak.EDU!toon%NEWS.SARA.NL
Sender: ANU-NEWS Discussion <ANU-NEWS@VM1.NoDak.EDU>
From: mcnc!VM1.NoDak.EDU!toon%NEWS.SARA.NL
Subject:      NEWGRPFILE/NEWITMFILE can create non-fragmented
              NEWS.GROUPS/NEWS.ITEMS
To: Multiple recipients of list ANU-NEWS <ANU-NEWS@NDSUVM1>
 
In article <9104232236.AA18459@jatz.aarnet.edu.au>, G.Huston@AARNET.EDU.AU write
s:
> Bill Glass sent me the following - I thought the response may be of
> interest to the full list:...
>
>       [Observations about GLOBAL_BUFFERS RMS parameter of news ISAM
>        files by Bill and Geoff deleted...]
>
> HOWEVER newitmfile and newgrpfile are not really very "nice" programs - the
> file they create is created using a very minimal set of RMS parameters -
> they are fragmented and totally untuned, as  they are the result of a heap
> of $PUT operations. (well not quite totally untuned as the $PUTS are in
> ascending order of primary key)
 
Well, they don't HAVE to be fragmented. I just added the following lines
to newgrpfile (and the corresponding code to newitmfile ...):
 
grpfab_1.fab$l_alq = 3*grpfab.fab$l_alq/4;/* Use 75 % of original allocation */
  ...
grpfab_1.fab$w_deq = grpfab.fab$l_alq/8;  /* and a reasonable extend qty     */
  ...
while ( !((status = sys$create(&grpfab_1)) & 1)
       && grpfab_1.fab$l_fop & FAB$M_CTG) {
  printf("Can't create %s contiguous (%d blocks); trying contiguous-best-try.
      grpfab_1.fab$l_fna, grpfab_1.fab$l_alq);
  grpfab_1.fab$l_fop &= ~FAB$M_CTG;
  grpfab_1.fab$l_fop |= FAB$M_CBT;
}
  ...
In words: Try to allocate three quarters of the original file size in advance;
first try Contiguous, then (if it fails) Contiguous-Best-Try. Works for me ...
--
 
Toon Moene, SARA - Amsterdam (NL)
Internet: TOON@SARA.NL
 
/usr/lib/sendmail.cf: Do.:%@!=/