- From: Phillip Hallam-Baker <hallam@ai.mit.edu>
- Date: Mon, 15 Sep 1997 15:49:40 -0400
- To: "'David W. Morris'" <dwm@xpasc.com>, Ben Laurie <ben@algroup.co.uk>
- Cc: Larry Masinter <masinter@parc.xerox.com>, "http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com" <http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com>, "http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com" <http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com>
>On Mon, 15 Sep 1997, Ben Laurie wrote: >> A proxy which attempts to buffer complete responses is broken, so it may >> as well be non-compliant. >Not if the proxy is required to respond to an HTTP/1.0 client with >chunked encoding converted to content-length. >Dave Morris In that case the proxy should buffer content as long as it can then send an HTTP/1.0 repsonse with no content length and slam the connection shut to denote end of transmission. This is how HTTP/1.0 servers deliver indefinite length encoded messages today. There is no reason to make support for http/1.0 with connection keep-alive any different. If there is a choice between resticted performance that does not break for legacy browsers and flexibility for the future it is almost always best to chose the latter. Phill
Received on Monday, 15 September 1997 12:57:22 UTC