- From: Roy T. Fielding <fielding@apache.org>
- Date: Tue, 20 Aug 2002 15:00:15 -0700
- To: "Larry Masinter" <LMM@acm.org>
- Cc: "'Jeffrey Mogul'" <Jeff.Mogul@hp.com>, "'Alex Rousskov'" <rousskov@measurement-factory.com>, "'Kim Horne'" <kim@pookzilla.com>, <ietf-http-wg@w3.org>
On Monday, August 19, 2002, at 09:43 PM, Larry Masinter wrote: > Roy, you're mixing server and client responsibilities. It's > quite possible to mandate "must not send a body with GET" for > clients and also mandate "must parse and ignore message bodies > with GET" for servers. Just the opposite. I was stating the requirement as it is currently written -- separating server and client responsibilities. The message parsing algorithm does not know about method semantics aside from HEAD responses. Whether or not a body is included with GET is not determined by the method semantics -- a server that does not anticipate and correctly parse a GET request with a body is not compliant with the protocol. Adding yet another requirement to HTTP just to say that a client MUST NOT send a body with GET, HEAD, DELETE, etc., will not change the interoperability of the current protocol and it absolutely must not change the message parsing algorithm (the example that Jeff was talking about). Doing so only eliminates the possibility that future extensions might use the request body for those methods. ....Roy
Received on Tuesday, 20 August 2002 18:00:46 UTC