RE: Documents, Cars, Hills, and Valleys

> one might grumble about, but all that aside much the same argument can
be
> run for 'classic' (CGI etc) Web services, ie. those that take a bunch
> of HTML-form POSTed parameters and return a (typically HTML or now
XML)

Actually, it *is* the exact same argument.  That is exactly what I was
talking about -- POST abuse.  POST abuse hides the resources behind an
opaque gateway that makes it impossible to identify anyone but the
gatekeeper.  And people get confused and start using the name of the
gatekeeper to identify the resources, since "it's the only endpoint we
have that we can talk to".  POST abuse severely erodes the value of URIs
as identifiers, and taken to the extreme makes URIs useless.

Like I said, we all agree that people do it; I am just saying that we
should acknowledge that there are people who really hate the fact and
want to make it end.

P.S.  Some would see SOAP as a way to segment things more cleanly
between "gatekeeper traffic" and "resource representation traffic", in
an attempt to reverse the erosion of URI sanctity.  That is, today it is
impossible to guess whether a use of POST is "proper" or not.  At least
SOAP makes it very clear that the message is not a "normal" POST, and
therefore all SOAP endpoints can be regarded as suspect.  And SOAP
provides a path of evolution that could lead to all of the RPC abuses of
POST moving away from POST altogether.

-J

Received on Wednesday, 10 April 2002 17:08:26 UTC