W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 2000

RE: On pipelining -Reply

From: Daniel Hellerstein <DANIELH@mailbox.econ.ag.gov>
Date: Fri, 21 Jan 2000 12:19:33 -0500
Message-Id: <s8884f2e.007@mailbox.econ.ag.gov>
To: lawrence@agranat.com
Cc: http-wg@hplb.hpl.hp.com
>If request1 is a GET, HEAD, TRACE, or OPTIONS, then I would say that
>the server could legitimately assume that it had no side effects.
>One could be more conservative and assume that any CGI (or CGI-like)
>resource had side effects.  Having assumed that there were no side
>effects it might be reasonable to overlap the processing, if you
>have the buffer space to spare.

That's a reasonable notion.  However, in practice it might be a bit
complicated (i.e.; server side includes in GETs of html documents
might
cause changes in  "state" variables.).

It seems that the tradeoff is between speed and
being-nice-to-less-then-careful-clients.
For now, I'll stick with sequential resolution, and then 
consider how to detect circumstances where 
sequentiality can be loosened.
Received on Friday, 21 January 2000 17:21:00 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:33:35 EDT