[
Lists Home |
Date Index |
Thread Index
]
Mike Champion wrote:
>
> One thing that doesn't seem to have come up in this mega-thread is how REST relates to
> things like WebDAV (which adds a few more verbs to HTTP to make collections of resources,
> lock and unlock resources, add metadata to a resource, copy and move resources between
> collections).
Method extensions are okay in REST as long as they are *generic* to all
resources, not specific to a particular one as in RPC. New methods can
be viewed as optimizations of existing ones. They are optimizations both
in the sense of better performance and simpler software construction.
* http://www.xent.com/pipermail/fork/2001-August/003191.html
> ... Also, what about things like the tuple-space TAKE operator (sortof a LOCK,
> GET, DELETE transaction)?
That's been proposed as an extension also.
> I can see how to model these with REST, i.e.,a series of GET, POST, DELETE operations and
> some logic to test for the existence of lock resources or metadata. But does bundling
> these into sequences of operations that are performed as a transaction violate REST, or
> merely extend HTTP?
Don't understand that part of the question. If you're asking whether
inventing new methods violates REST, the answer is no. The only problem
is that deploying them is a headache.
Paul Prescod
|
