Jason, I hope you don't mind me cc-ing the quixote list. But I've
desperately searched the archive for an answer to this. Finally there,
will be one now.

Jason wrote:
> Gerhard Haering wrote:
>>
>> Sure, threads can be quite a hassle. But an application server that
>> canot handle more than one concurrent request doesn't sound like a
>> good idea to me, either. For anything but low-profile and toy web
>> projects, that's not acceptable. From the standpoint of the web
>> framework user threading will happen transparently, anyway.
>
> This is factually inaccurate. Medusa can handle more than one request at
> a time, it just uses asynchronous I/O and select to multiplex that
> within a single thread. The question of whether this is a net
> performance gain or loss is, like most performance questions, something
> that usually comes down to: "it depends"; although it turns out to be a
> win far more often than you might think.

I know a little about the asynchronous model as I've studied Twisted
lately. Still I suppose a Quixote request is processed in one go
(Medusa/Twisted can't suspend the processing of one request to accept
the next; doing so requires a callback-based API), so I'm not sure how
using Medusa/Twisted had any advantage here over a plain stupid solution
using Python's HTTPServer.

>> My question was if SCGI for example uses threading internally or if
>> I'm stuck with one (1) concurrent request when using SCGI as a Quixote
>> backend? How about mod_python?
>>
>> Anybody can answer this, pleaaaase?
>
> SCGI uses fd passing to pass off the work to children processes (by
> default a max. of 5).

That's very good to know. Thank you for this information.

-- Gerhard