Hi Sanjiva,

On 11/22/07, Sanjiva Weerawarana <[EMAIL PROTECTED]> wrote:
> I gave a talk on ws-* vs. rest .. once at QCon and then at ApacheCon last
> week. See: http://wso2.org/library/2818
>
> Comments welcome!

Where to begin? 8-)

You seem to go back and forth between describing REST as an
architectural and as shorthand for "using the Web".  For example, you
say "Lie: REST is easy to learn", but then you list a bunch of specs,
some of which are used in developing Web based software.  Whether REST
is easy to learn or not has nothing to do with those specs of course,
because REST is just the style.  You also say "Myth: REST is
multiprotocol" which suffers from the same problem.

You say "Myth: HTTP has a uniform interface", and then present WebDAV
& PATCH as evidence that it doesn't.  In fact, all those extension
methods are uniform because they can be invoked on all resources.
"Uniform" doesn't mean or imply "non-extensible".

You also say "True REST is still an art form" and present as evidence
the length of time it took to develop APP.  Of course, it would take
100 people on any WG that long to agree on what to have for lunch.
Web3S - which I contributed to, BTW - has its own issues and is less
RESTful than APP because it doesn't use
hypermedia-as-the-engine-of-application-state exclusively.  Anyhow, my
experience with it has shown me there's no fuzziness or room for
interpretation with REST and that its definition is as tight as any
I've worked with.

It's not a myth, RESTful systems do not need WSDL.  It doesn't need
any description languages in fact.  It needs forms: information sent
by the server which instructs clients on their options for progressing
through the application state machine.  At runtime.  Any
development-time dependency that one software component might have on
another - beyond those defined by open standards of course - is an
implementation dependency that serves to tightly couple them.

I don't see what security & caching have to do with scalability.  Are
you talking about REST or HTTPS?

I don't follow the "Lie: RESTful services have a uniform interface"
argument, but it's prima facie incorrect because they do by
definition.

Cheers,

Mark.
-- 
Mark Baker.  Ottawa, Ontario, CANADA.         http://www.markbaker.ca
Coactus; Web-inspired integration strategies  http://www.coactus.com

Reply via email to