Re: WSDL Debate Continued..

Bijan Parsia wrote:

> Small point: There is no WSDL debate in any meaningful sense. We've been 
> trolled (as Drew's post makes evident). Sorry for making it seem otherwise.
> 
> Frankly, just the nature of the W3C makes it nigh impossible to 
> seriously depart from WSDL without an overwhelming rationale obviously 
> convincing to all. I'm hard pressed to imagine what that could even been.
> 
> Given that *three* recent WS submissions (transfer, eventing, and 
> enumeration) all build/rely on WSDL (and Addressing, which is connected 
> to WSDL), suggests that, at least at the moment, enhancing WSDL is the 
> way to enrich WS descriptions. So that's how we'll go.
> 
> (Where are the SWS-* specs? Let's start with preconditions and 
> effects...my personal favorite. Isn't that hard! Or non-functional 
> properties. Dublin core anyone? Again easy. We don't need a working 
> group to make a proposal! Anyone interested? (Also, don't focus on the 
> RDF mapping...focus on the normal WSDL (*provide* an extended mapping, 
> natch).)
> 
> (This is one reason I don't think WSDL-S gets us much. We don't need 
> *hooks* we need substantive content. WSDL is hooky enough.)

I agree that a set of new hooks (as exemplified by WSDL-S) probably 
won't get us very far - not by itself, that is.  Of course the WSDL-S 
vision allows the hooks to point to anything (OWL-S, UML, WSMO, ...), 
which could be useful, but I think the community is also going to want 
some standardization regarding what's pointed to (and the implications 
of the pointing) before it gets really useful.

One very natural (and W3C-oriented) way to go, seems to me, would be a 
simple 2-step approach:

(1) Go ahead and standardize something like WSDL-S (as envisioned in the 
sa-ws charter: http://www.w3.org/2005/10/sa-ws-charter).

(2) Get the SWS folks together with interested WS stakeholders in 
another working group to spell out how to specify, in a SemWeb-based 
manner, the low-hanging fruit that's at the intersection of most of the 
SWS proposals.

In my mind, what you mentioned above - preconditions, effects, 
non-functional properties - is a large part of this low-hanging fruit. 
Another piece is the ability to use a SemWeb language to specify the 
types of inputs and outputs (something which, in principle at least, you 
can already do in WSDL).  Another likely piece is the ability to locate 
services within a class hierarchy, around which your non-functional 
properties would be organized, and extended into specific domains.

Of course, once you've done (2), then it's not that important whether 
the resulting specs are located directly within a WSDL document or are 
pointed to, in some external location, by the WSDL-S annotations (as 
several people have observed).

This second step has been discussed in various conversations among the 
SWS camps, and I (for one) have the sense that it is pretty readily 
doable.  In fact, Rick Hull proposed more-or-less these 2 steps at the 
W3C Workshop on Frameworks for Semantics in Web Services.  Unfortunately 
at that point there wasn't adequate time left for discussion but I think 
many of the SWS folks, at least, were in sympathy with the idea.

- David

Received on Saturday, 18 March 2006 00:42:17 UTC