Fwd: Support for GZipped-Batches in sendbatch?
davidsen at tmr.com
Wed Apr 25 19:20:32 UTC 2001
In article <yl8zkshvl6.fsf at windlord.stanford.edu>,
Russ Allbery <rra at stanford.edu> wrote:
| Forwarded to inn-workers from inn-suggest. Do any of the folks who use
| UUCP have any comments on this?
| From: Jochen Erwied <jochen at erwied.de>
| To: inn-suggest at isc.org
| Subject: Support for GZipped-Batches in sendbatch?
| Date: Sun, 30 Jul 2000 14:45:14 +0200
| There seems to be support for compress only when sending out batches with
| sendbatch. I added support using gzip (options -g/+g). Maybe one should
| think about using bzip2, too. But this is far from being standard.
| Relevant part in sendbatch is:
[ snip ]
| where GZIPFLAGS contains a '-9f'.
Arrrgh! That one I do suggest should not be default! The gain from level
7 to level 9 is rarely more than a few percent, and the CPU time
increase is often 2-4x. I think gzip is a fine thing to add, but I think
the default should be something more conservative. Actually, for
binaries the gain drops off at about 3, but the cost doesn't change much
until 8 and 9.
| By the way: COMPFLAGS should be set to '-f' by default!
That should go in STABLE, it's a bug.
| Comments welcome.
A good thing to do, and well worth the trivial effort to do it. A big
gain when feeding a low bandwidth site.
Note for those who haven't done UUCP over TCP: this is really good in
terms of bandwidth saved. In addition to the compression, you get fewer
interractions, so the limit becomes bandwidth (between Pentium 200's you
can easily max a T1). And since the connection is originated when the
leaf site wants, it's easy to run over a line used for "real work"
during the day, taking only time critical stuff over nntp, then feeding
the rest via UUCP in the low load hours.
bill davidsen <davidsen at tmr.com>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
More information about the inn-workers