Re: Should Web Services be served by a different HTTP n+1?

On 2013-01-24 04:18, Grahame Grieve wrote:
> Hi Tim
>
>> I disagree strenuously.  The semantics of HTTP as they are today map well
>> onto my RESTful-app-construction needs.
>
> well, here's an issue I have. What error code should an http server return if
> a proposed update is not suitable because of the content of the submission?
> There's nothing wrong at the http layer - it passes authenticaion, currency
> and content type checks, and parses, ok. But there's something in the content
> that violates server business rules. Something like trying to submit and
> order with an identifier that has already been used, for example.
>
> What would be right http status code to use? It's a client error, right?
> The nearest appropriate status code would be 422, but I'm not sure
> whether that can be used outside webdav. Either way, there's a bunch

It can.

> of things I would expect in an application exchange protocol that don't
> map that well.
>
> I don't think that a different protocol is a good idea, but the discussion of
> what's planned for HTTP/2 seems very focused on browsers & the web to
> me, and doesn't seem to me to adequately consider the huge use that
> exists for application/application communications

Augmenting error handling for web services is an interesting topic. See 
prior proposals about Link relations, or a JSON typed response body 
format for 4xx/5xx.

All of this can be done in HTTP/1.1 and doesn't require any protocol 
changes.

Best regards, Julian

Received on Thursday, 24 January 2013 15:30:35 UTC