>>>>> "BS" == Bob Sutterfield <Bob@XC.Org> writes:
BS> Has anyone compiled MajorDomo 1.94.3 with Perl 5.004_1 and the Perl
BS> Compiler Kit a3? Did it work?
It should mostly work; the a4 compiler (which might not be in general
release) is probably better but no compiler works at all with perl 5.004
(according to Malcolm) so you're hosed anyway.
BS> Any improvement?
You wouldn't really get much; the compiler mostly helps with startup time
and much with execution time, at least for the kinds of things that
Majordomo's doing.
BS> Almost all of our 1432 lists use restrict_post, so we're very sensitive
BS> to resend speedups too, as observed in
BS> http://www.cis.ohio-state.edu/~barr/majordomo-faq.html#4.5.
If existence testing is what's getting you, you'll have to wait until a
general database access mechanism comes into use; this won't be until 2.0
although some good hacks are possible with only a few hours of development
time[1]. This will enable O(log n) lookups instead of the current O(n)
lookups. If compilation time is getting you, then 2.0 will be even worse
(much more code to compile) but there are design plans for a persistent
server in a far future version.
BS> And we depend on some features we've added to sendmail 8.8.6, so
BS> http://www.qmail.org isn't really an option either.
Are you interested in processing speed or delivery speed? The
qmail/Sendmail decision doesn't affect how long it takes Majordomo to do
whatever with the message and hand it to the MTA.
[1] Maintain a database _in addition_ to the list file; have the
subscription routines alter it and have is_list_member check it. Should
take just a few hours of hacking and testing for someone who knows the
internals. There are complications I haven't considered (mungedomain for
one).
- J<
References:
|
|
