On Sun, 27 Sep 1998, Roger B.A. Klorese wrote:
: On Sat, 26 Sep 1998, Bill Houle wrote:
: > Roger B.A. Klorese said:
: > > It's cart-and-horse time. Maybe we ought to be positioning Majorcool as
: > > the product, and Majordomo as a pile of back-end software it uses.
: > For a Web-oriented site, this may sound like a good idea. But I have
: > been in (or listened to) many arguments about keeping the MLM field
: > Web-free, or at least, keeping it non-dependent upon the Web. There
: > is something to be said for the least-common-denominator appeal of
: > email. A "Web first" approach may be a year or so too soon.
:
: I'm speaking of *positioning* and not *design*. I've worked for the past
: seven years on a product which consists of a couple of dozen commands and
: drivers, plus a GUI (once Motif, now Java). We originally documented
: and positioned it as a command-line tool which happens to have a GUI. It
: took off in acceptance, and vastly increased in user understanding, when
: we positioned the GUI as the primary interface, then described the CLI as
: an advanced or low-speed-access alternative.
I think mj2 is going along way to just what you describe. The new mj_*
command interface make a GUI much much much easier then before.
However I stongly believe that mj2 needs a strong compact core. If designed
well, (as it appears to be going to me) it will have the hooks to interface
to a number of services.
I did not mention last week, but there was this interesting *performance*
page on http://www.lsoft.com/listserv-perf.html comparing how glaringly slow
the original vanilla majordomo is compaired to listserv. Now you know for a
fact that the tests were not run with anything like Bulk_Mailer included, so
I took it with a little grain of salt. However I have compaired my mj1
server to my mj2 server for a very small list, and the mj2 baseline is quite
a bit faster, which is a nice thing. However I am curious how it scales with
large lists. One thought I have had was instead of doing the bulk mailer
equivalent at deliver time, that it might be better to sort the addresses at
subscription time. (Handwave time ... not sure what Jason's replacement is
doing). A database indexed by MX records might make for a swifter sending
phase also.
Thoughts on performance for large lists?
Randall
Follow-Ups:
-
Performance
From: Jason L Tibbitts III <tibbs@hpc.uh.edu>
References:
|
|
