Quinn Dunkan writes:
> Well, if SCGI works as a plain CGI script (connecting
> to a long-running python) then PCGI would have little
> advantage.  Last time I messed with Zope (quite a long
> time ago) they were talking about mod_pcgi, so maybe
> that's been done by now.

Andrew Kuchling wrote an implementation of mod_pcgi,
and then Oleg BroytMann developed another version,
mod_pcgi2:

http://phd.pp.ru/Software/Zope/

Oleg's motivation was support for the cryllic alphabet,
which wouldn't work with FastCGI.

> PCGI has been around for a while, and I assume it's
> been banged on more than SCGI, so there could be a
> stability win.  And just thinking of the whole
> "no sense reinventing existing code" thing.

I'm surprised PCGI has been around as long as it
has.  It's a very simpleminded (and flawed) protocol.
I proposed some changes back a few years ago, but
they were put on the backburner.  Since then, I've
no longer been doing any active work with PCGI,
though we still use it on our servers.

an-actively-supported-SCGI-trumps-a-dormant-PCGI-ly y'rs,

Jeff Bauer
Rubicon Research