RE: Belated comments on SPARQL Protocol for RDF 25 January 2006 LC WD

On Wed, 2006-03-22 at 10:33 -0500, Arthur Ryman wrote:
> Dan, 

> BTW, there is a WSDL extension attribute wsdlx:safe [1] You can use
> that to mark on operation that is known to be safe, in which case it
> would be bound to GET. 
> [1] 

I assumed we were using that. Kendall, are we not?

> Is the query operation supposed to be safe?  

> <operation name="query"
> pattern=""> 
> You have two HTTP bindings for it, GET and POST. Doesn't this violate
> Web architecture?


>  If the operation is safe, it should be bound to GET.

It _is_ bound to GET; it's also bound to POST.

>  If the operation is not safe, it should not be bound to GET. Seems
> like binding the same operation to both GET and POST should never
> happen. Or are you leaving it up to the user?

Yes, with this advice:

The queryHttpGet binding should be used except in cases where the
URL-encoded query exceeds practicable limits, in which case the
queryHttpPost binding should be used.

See also section 5 Practical Considerations in the relevant TAG finding:

>  Does the safety depend on the actual query, i.e. some querys do
> updates (I have read the SPARQL spec) ?

We haven't specified updates in this version; nonetheless, some
queries don't fit in GET.

>  If so, having both bindings does make sense.Thx. 

> Arthur Ryman,
> IBM Software Group, Rational Division
> blog:

Dan Connolly, W3C
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Wednesday, 22 March 2006 16:34:43 UTC