Hello,

Dieter Maurer wrote:

> Maybe, I have a much simpler solution:
>
>   Something in ZServer makes all file fields seekable by
>   delivering them through some temporary (either a StringIO
>   or a temporary file).
>
>   Maybe, you could do the same for your requests?

Yes, I could.  But trying to avoid it was the very reason I tried to
patch HTTPRequest for.

What you propose is what the HTTPServer does for instance.  I have at
least two objections:

1) For big requests (big means some hardcoded threshold) HTTPServer
   buffers the request on disk.  Then ZPublisher creates a
   FieldStorage instance which will write the request once again to
   disk.  And that also means the request is read from disk several
   times.  That's a bit too much blocking disk I/O in a production
   webserver for my liking.  And that's not counting retries; each
   retry will add another pair of blocking disk I/O in the current
   implementation.  What I'm trying to do is:

   - avoid unneeded disk I/O in retries

   - make it possible that self.stdin in HTTPRequest can be
     non-seekable such that it doesn't have to be buffered on disk
     before FieldStorage is created (and does exactly that buffering).

2) I'm convinced that the one and only ZServer thread should do
   without any blocking (I/O) call (besides select() that is) and I
   want AJPServer also be an experiment to see if this improves
   performance.


cheers,

andreas

_______________________________________________
Zope-Dev maillist  -  [EMAIL PROTECTED]
http://mail.zope.org/mailman/listinfo/zope-dev
**  No cross posts or HTML encoding!  **
(Related lists -
 http://mail.zope.org/mailman/listinfo/zope-announce
 http://mail.zope.org/mailman/listinfo/zope )

Reply via email to