Tres Seaver wrote:
Hash: SHA1

Jim Fulton wrote:

Jim Fulton wrote:

Tres Seaver wrote:


Ob. note:  the performance characteristics of such servers (including
twisted) are not well understood in the context of Zope, until some
brave soul actually rolls out a high-volume production site and reports
success or failure.

Good point!  It would be good to investigate this issue further.

I did some investigation and found the thread-management strategy
currently provided to be unsatisfactory.

Given the twisted team's general attitude toward threaded programming, I
don't find that terribly surprising.  Kind of like what you get when you
elect government officials whose sincere program and belief is that
government can't work. ;)


FWIW, I'd really like the threading strategy to be pluggable at the
WSGI level.  I'm trying to push that, but I'm getting some resistance
because this is really only an issue for async servers. OTOH, I
like async servers.

I've posted a note to
the Twisted list.  I don't know if this can be fixed before the
next Zope 3.2 beta this weekend.

I think I'm going to look into adding an option to makezopeinstance
to select which server is used.  I haven't decided yet which one will
be the default.  I'm leaning toward making twisted the default to
try to increase the amount of testing it gets, but recommend that
people use ZServer for now in production sites.

- -0.  I would restate my belief that the twisted integration is
yet-to-be-proved in "production" use, and needs some care-and-feeding
from people whose pagers go off (or the equivalent) if it breaks.

I would restate my agreement.

> I can
live with the default being twisted, but it is a bit like leaving
'debug-mode on' in the default config file for now.



Jim Fulton           mailto:[EMAIL PROTECTED]       Python Powered!
CTO                  (540) 361-1714  
Zope Corporation
Zope3-dev mailing list

Reply via email to