[2402] in SIPB bug reports
GRIPE about XRN 6.17
daemon@ATHENA.MIT.EDU (Jonathan I. Kamens)
Sun Jan 26 14:21:27 1992
Date: Sun, 26 Jan 92 14:20:49 -0500
From: "Jonathan I. Kamens" <jik@pit-manager.MIT.EDU>
To: prc@Athena.MIT.EDU
Cc: bug-sipb@Athena.MIT.EDU
In-Reply-To: prc@Athena.MIT.EDU's message of Fri, 24 Jan 92 18:08:05 -0500 <9201242308.AA12876@M11-116-13.MIT.EDU>
From: prc@Athena.MIT.EDU
Date: Fri, 24 Jan 92 18:08:05 -0500
I was running xrn and rn in parallel. On entering a
newsgroup with 16 unread articles from both programs
simultaneously, I was able to read 14 of the articles using rn
before xrn brought up the first one. I know xrn must be doing a
lot of I/O to get the article header information, but it sure seems
slow. Since I read all the articles, rn must have done the same
(or more) file I/O, plus it had to wait on me!
Rn did *not* do "the same (or more) file I/O" as xrn; in fact, xrn is
doing quite a bit more than rn, which is why it is slower.
Unless you ask rn to give you lists of article headers with something
like the "=" command, then only interactions it has with the NNTP
server is asking for entire articles. The NNTP server is capable of
delivering entire articles more quickly than it is capable of
returning lists of header lines from ranges of articles. Xrn asks it
to do the latter in order to build up the list of subject lines,
authors and article sizes that it displays when in a newsgroup.
So xrn is doing more I/O with the server than rn is, and it's doing
operations that are slower on the server for various reasons. It is
not surprising that xrn is slow, because the news server is
overloaded.
We are getting a new disk drive for the news server with more space on
it, so that we will only have to run "expire" (which takes a lot of
CPU time) once per day, late at night. Also, we are adding more RAM
to the machine that is the news server. Once these improvements are
complete, you should notice a significant improvement in performance,
and I suspect you will be quite satisfied with the resulting speed of
xrn.
--> Jonathan Kamens
IS/Athena Quality Assurance and User Consulting
Treasurer, Student Information Processing Board (SIPB)
jik@Athena.MIT.EDU
