🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: gnu.emacs.gnus
Article #871

Xref field vs cpu

#871
From: Robert D. Houk
Date: Fri, 09 Feb 1990 21:36
64 lines
3383 bytes
What exactly is the definition of the Xref field? I realize that it is a
"cross-reference" 'twixt groups for the same message, but does anyone have
an exact or "reference" definition?

The reason I ask is because GNUS (3.12, Emacs 18.54) oftimes goes into a
cpu-burning state on "q"uitting a newsgroup that has one or more messages
with an Xref field. The worst example was a brand-new alt.somethingorother
with 9 messages which took 75 seconds elapsed time on a Sun 3/80 - cpu
running flat out, no I/O, no network, no paging (from perfmeters). Has
anyone else seen this? Fixed this?

For example, the following message from alt.sys.sun takes 18 seconds
running flat out to "q"uit the alt.sys.sun group (with a total of 2
messages).

    Xref: sli alt.sys.sun:464 news.groups:15311
    Path: sli!uunet!shelby!apple!usc!chaph.usc.edu!aludra.usc.edu!gcrum
    From: gcrum@aludra.usc.edu (Gary Crum)
    Newsgroups: alt.sys.sun,news.groups
    Subject: Re: comp.sys.sun performance
    Message-ID: <GCRUM.90Feb7151901@aludra.usc.edu>
    Date: 7 Feb 90 23:19:01 GMT
    References: <15966@well.UUCP> <4579@brazos.Rice.edu>
    Sender: news@chaph.usc.edu
    Followup-To: alt.sys.sun
    Organization: University of Southern California
    Lines: 32
    In-reply-to: rgreene@wild.rice.EDU's message of 5 Feb 90 20:54:06 GMT

    [Body of message deleted]

If I select only 1 message without the Xref, then "q"uitting is virtually
instantaneous, as is "q"uitting (e.g.,) comp.sys.sun with 161 Xref-less
messages.

The above times are very consistent and reproducible. The problem is
also reproducible on our Sequent system (same GNUS/emacs).

It seems to make no difference whether or not the Xref'ed groups are
subscribed or not. It seems to make no statistically-significant dif-
ference how many groups are Xref'ed. Most aggravatingly of all, it does
not always happen - some Xref'ed message/groups still "q"uit quickly.
It doesn't seem to matter how many groups are in my .newsrc (I tried
c-K'ing most of the unsubscribed groups out). No "obvious" difference.

Related problem, er, observation: GNUS still marks individual Xref'ed
messages are read even in the unsubscribed groups. Over half of my
.newsrc (and of course .newsrc.el) file(s) consisted of comma-separated
individual message numbers in unsubscribed groups! Any way to suppress
this?

It's been a few years since I almost learned Lisp, so I'm hoping someone
out there has already solved my problem . . . any help greatly appreciated
retroactively in advance!

					-RDH

--
  "If you build a system                 RDH@SLI.COM  |	 uunet!sli!rdh
 that even a fool could use,             Robert D. Houk
   then only a fool                      Software Leverage, Inc.
 would want to use your system."         485 Massachusetts Avenue
        -Somebody-r-Other's Law          Arlington, Mass.  02174
                                         Office (617) 648-1414
Technical Details
Message-ID: <RDH.90Feb9163114@borodin.sli.com>
Path: arc1-ca.pugleaf.net!utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!sli!news

Show Full Headers
Complete Article Headers
Path: arc1-ca.pugleaf.net!utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!sli!news
From: rdh@sli.com (Robert D. Houk)
Newsgroups: gnu.emacs.gnus
Subject: Xref field vs cpu
Message-ID: <RDH.90Feb9163114@borodin.sli.com>
Date: 9 Feb 90 21:36:25 GMT
Sender: news@sli.com (news)
Reply-To: rdh@sli.com (or uunet!sli!rdh)
Distribution: gnu
Organization: Software Leverage, Inc. Arlington, Ma
Lines: 64