Lists Home |
Date Index |
On Friday 08 February 2002 06:03 am, Paul Prescod wrote:
> > FWIW. I wrote a distirbuted hypermedia/collaborative system back
> > in the late 80's where everything used s-expressions. That system
> > only had GET, SET and CALL... DELETE was (set foo nil)....
> You can certainly build a distributed system in the manner you did.
> But I would suggest that DELETE is useful because HTTP does not try
> to interpret your message content. It reserves the message body
> entirely for the application. You were *building* an application so
> you made a different decision.
Oh I agree. DELETE is useful... overloading SET was a bit of a kludge.
I was just pointing out that you need very little to do a lot... the
trick in the applications use of the protocol, not the protocol itself.
> Unlike newer specs, most of that complexity was added by or in
> consultation with, implementors, as they were implementing, to solve
> problems they found in implementation.
I'd go back and review this though... things like content negotiation
could arguably be eliminated, or at least trimmed down a bit. HTTP 1.1
is not a simple protocol.... and some parts were added without much
thought (like Accept-Charset ;-))
> We're not going to get back into "what is HTTP for?" are we? ;)
No. My feeling is that the *model* you're espousing could be built on
a cleaner substrate. That's my opinion, just as it's *your* opinion
that HTTP is all you need.