[11491] in www-talk@info.cern.ch
Re: Multiple responses from one request
daemon@ATHENA.MIT.EDU (Patrick McManus)
Thu Feb 17 10:31:29 2000
From: Patrick McManus <mcmanus@appliedtheory.com>
Message-Id: <200002171524.KAA09453@justice.atc-bos.com>
To: DANIELH@mailbox.econ.ag.gov (Daniel Hellerstein)
Date: Thu, 17 Feb 2000 10:24:32 -0500 (EST)
Cc: www-talk@w3.org
Reply-To: mcmanus@appliedtheory.com
In-Reply-To: <s8abc81e.008@mailbox.econ.ag.gov> from "Daniel Hellerstein" at Feb 17, 2000 10:04:17 am
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Resent-From: www-talk@w3.org
In a previous episode Daniel Hellerstein said...
::
:: I suspect this is a no-no, but my reading of rfc2616 doesn't give me a
:: definitive answer;
::
:: The question:
:: Can an http 1.1 serverr "pipeline" multiple responses (over a
:: maintained connection)
:: to a single request; without using a mulitpart content type (and
:: suitable dividers).
::
you're right.. a definite non-starter.. 1.4 makes it pretty clear that
there's a single response for every request..
consider in a a persistent scenario if 1+ responses where allowed for
each request, and the browser sent a second request to the server (it
is persistent..).. how does the browser know if response 2 is tied to
the first or second request? (and remember that the server may not
have even received the 2nd request even though it's entirely out of
the browser..).. it's a labelling issue.. it's one of the same reasons that
you can't return responses in a different order than you received the
requests, even if head of line blocking makes that inefficient.
-P
:: For example
:: a) client requests /pictures/anim5.cgi
:: b) the "anim5.cgi" script responds with 5 differerent images, say
:: with a second or two wait between images
:: c) This response is over the single connection
::
:: Thus, the response would look something like;
::
:: http/1.1 200 Ok
:: Content-type: image/gif
:: Content-length: 251
::
:: (image1 info, that is 251 bytes long)
:: http/1.1 200 Ok
:: Content-type: image/gif
:: Content-length: 412
::
:: (image21 info, that is 412 bytes long)
:: ....
:: http/1.1 200 Ok
:: Content-type: image/gif
:: Content-length: 1311
::
:: (image5 info, that is 1311 bytes long)
::
:: Okay, this may be a stupid thing to do, given the way most browsers
:: will react
:: (i.e.; they won't replace images to form a server push animation)..
:: The point is that I'm trying to decide if I should attempt to
:: detect, and then disallow, this sequence of events in a "customizable"
:: http/1.1 server I'm revamping.
::
:: Thanks
::
:: Daniel Hellerstein
:: http://www.srehttp.org
:: danielh@crosslink.net
::
::
