Re: OldeWorldeWebServices?

Hi Dan,

You mention in there that you use - and rightly so, IMO - URIs to denote
the things in that space, be they Quake servers or whatever.  That's
great.  But have you considered that information like "port number" is
typically part of the URI, and that some of the other information you
associate with a service description like this is also implicit (i.e.
there already exists a publically specified means to get there from the
URI, such as from URI scheme -> protocol)?

Consider your opac1 z39.50 example.  The descriptive information includes
the following;

- protocol name, which can be looked up via URI scheme
- protocol spec URI, which could be specified as part of the
  registration of the scheme I suppose
- path, which is already part of the URI
- host, which is already part of the URI
- port, which is either part of the URI, or registered with IANA
- operationType - not exactly sure what you're trying to do with that, but
  wouldn't the URI scheme suffice?

FWIW, I don't think it's a coincidence that the information you need
to describe these types of services is all available either directly
or indirectly via a URI.  Indeed, the decentralized & late-bound
interaction style that all Internet scale systems use (of those I've
looked at, anyhow) requires it, as to be missing a piece of information
requires a separate (potentially - and likely in practice - centralized)
authority be available to provide it.  I wrote a little about that here;

http://www.oreillynet.com/pub/wlg/1680

MB

On Mon, Apr 14, 2003 at 03:09:10PM -0400, Dan Brickley wrote:
> 
> Hi all
> 
> I had an idea for a Web services-ish spec a while back, finally wrote it up. 
> 
> Feedback/suggestions welcomed. I scribbled it in a Wiki so feel free to 
> annotate it inline if you like.
> 
> The idea is to use markup to describe pre-XML Internet-accessible services,
> using minimalistic and namespace-extensible markup giving port, host, 
> protocol, etc data., plus whatever else makes sense using other namespaces.
> It may well be that WSDL or other existing WS specs can do this already, 
> although my understanding was that they focussed on XML (esp SOAP) 
> based services.
> 
> The draft sketches a way of making Z39.50, LDAP, Quake or whatever 
> servers findable via XML descriptions. I use RDF though that's not 
> intrinsic to the general approach.
> 
> Anyway, http://esw.w3.org/topic/OldeWorldeWebServices
> 
> thanks for any comments,
> 
> cheers,
> 
> Dan

-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca
Web architecture consulting, technical reports, evaluation & analysis

Received on Monday, 14 April 2003 16:15:00 UTC