W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2007

Re: New issue: Need for an HTTP request method registry

From: Henrik Nordstrom <henrik@henriknordstrom.net>
Date: Tue, 07 Aug 2007 12:07:52 +0200
To: Adrien de Croy <adrien@qbik.com>
Cc: HTTP Working Group <ietf-http-wg@w3.org>
Message-Id: <1186481272.19935.13.camel@henriknordstrom.net>
On tis, 2007-08-07 at 16:25 +1200, Adrien de Croy wrote:

> And what does it all mean for caching?

Not much. There is already a catch-all phrase defining how caches should
operate when seeing unknown methods. All cached entities of the
request-URI is invalideted.

This phrase needs to be refined a bit to avoid unneeded cache
invalidations, but that's an optimization.

> Another example of the perpetual tug of war between complexity and 
> conformity I guess.

And is why the set of REQUIRED methods in HTTP should be kept down to a
minimum, with most of the methods only required in their context. i.e.
WebDAV methods only required in clients and servers supporting WebDAV.

> I agree we shouldn't let broken implementations define HTTP - that's the 
> tail wagging the dog.  I do think we should however always keep in mind 
> implementation difficulties when designing protocols.

Yes.

> Sometimes it 
> appears that bug-free implementation has been considered a given where 
> in fact it's non-trivial.  Unfortunately it's us fallible humans that 
> keep implementing things, and mistakes will always happen, and we will 
> always be lazy and try and implement as little of a spec as we 
> absolutely need (whether or not we can truly perceive how much that 
> is).

Sure, and equally true no matter what level things gets implemented at.

This said it's not very hard to extend HTTP without breaking the core.
But some aspects of the core is not understood well enought currently
due to the complexity of the specification and this is something that
needs to be addressed. In itself the features is not complex.

- The self-contained message nature of the protocol and transport
independence (violated by NTLM/Negotiate).

- Cache model with Request-URI, ETag, Content-Location, as seen in
recent discussions. (problem for WebDAV PROPFIND, but not strictly a
violation)

- ETag in general.

> Humans seem to have a tendency to complexify things over time 
> rather than simplify them - it's a common evolutionary process,

Commitees have that tendency yes. But generally not individuals trying
to make their life. And with proper focus comitees can occationally
simplify things..

Regards
Henrik

Received on Tuesday, 7 August 2007 10:08:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:50:15 GMT