Re: Persistent connections: current implementations?

Larry Masinter wrote:
> Specific feedback on the HTTP/1.1 specification from the point of view
> of either "is it easy to understand" or "what parts are difficult to
> implement" would be really useful at this point. We're extremely
> interested in the experience of implementors.

I can't say that I've had any specific problems with the spec, but I read
it back in November and there's been Christmas in between so maybe the
Xmas BBQs and beer have clouded my memory ;-)

Getting the persistent connections right has been tricky, but that's more
to do with the code I've inherited than being caused by HTTP/1.1.

I'm only part-way through implementing the various MUSTs, so I may send
further comments in the future.

> I'm especially interested in your reaction to the various "MUST"s and 
> "SHOULD"s.

I think it's very useful to have that distinction - to know what you can rely 
on when a software advertises itself as being HTTP/1.1 compliant, and being
able to define a test suite.  However, it does make incrementally implementing
the software difficult, since it's all-or-nothing.  For example, my server
calls itself HTTP/1.1 compliant - it does do persistent connections by
default - but it is not since it does not yet understand chunked transfer
encodings.

Regards,
Steve Ball