Re: [rtcweb] Fwd: New Version Notification for draft-uberti-rtcweb-turn-rest-00.txt

[Adam Roach <[e-mail]>]

On 7/8/13 16:57, Martin Thomson wrote:
> On 8 July 2013 14:50, Adam Roach <[e-mail]> wrote:
>> I only find one issue in my initial read through the document: why are we
>> proposing the use of HTTP POST for stateless, idempotent information
>> retrieval? The key benefit of this approach is that it doesn't actually
>> create state in the network. That's a much better fit for GET.
> That's an interesting question.  From one perspective, this is not
> idempotent, nor is it safe.  It creates a new resource: a new
> username/password pair with a defined validity.

That's like claiming that every time I hit cnn.com, it creates a new 
resource, since the organization of the articles and ads changes as a 
function of time (new articles are added) and local cookies (encoding my 
section preferences). I think both are kind of a tortured interpretation 
of resource creation.


> But it's equally valid to try to pretend otherwise and use GET.  After
> all, most servers won't actually be allocating any state for these
> things.

I hope none of them do. It's kind of the main point of the whole draft.

> p.s., I was going to let this slide, but this design is not RESTful,
> no sense in pretending like it is.

I'm not sure I agree with your assessment. Which of these six 
constraints do you believe are violated?

http://en.wikipedia.org/wiki/Representational_state_transfer#Constraints

/a