--- Sanjiva Weerawarana <[EMAIL PROTECTED]> wrote:

> True but there at least have been several attempts: ADS
> (Advertisement
> and Discovery of Services), WS-Discovery, WSIL (Web Services
> Inspection
> Language), and of course WS-MetadataExchange. 
> 
> The last is not a discovery protocol as much as it is a way to get
> metadata once you know the endpoint of a service. In the REST space
> none
> of this stuff existed because REST has been used for human driven
> apps
> where its not needed .. google works just fine.

Last I checked, HTTP exchanged metadata just fine, and even allowed for
negotiation (e.g. SPNEGO) without any human intervention. 

   WS-MetadataExchange even optionally piggybacks on WS-Transfer (!!!)
which is all about applying REST contraints to a SOAP world.  

> > Word doc and email it to your colleagues, or do you use UDDI or
> maybe  
> > a vendor-specific or homegrown registry, or perhaps you use yet  
> > another approach that I haven't mentioned?
> 
> You have to agree that at least in WS-* land there exists WSDL and a
> bunch of efforts to solve the discovery problem.

None of which (other than WSDL) are standards or widely implemented. 

All of these efforts to solve the discovery problem strike me as
inferior to REST itself, i.e. a system that describes the interfaces in
hypermedia -- and in fact that's what most people do with web services
interfaces like WSDL anyway - they make them web accessible, either
through the engine, or through a registry's web interface.

One reason Systinet's UDDI repository became so popular was that it had
a great web interface, with well-defined URIs for accessing its
resources.  I don't think that's a coincendence.

  
> In REST land until recently restafarians said "ah WSDL-like ugly
> static
> binding mechanisms are not needed; we're so hang loose types" but
> have
> now finally come to the realization that it is needed: Enter WADL,
> URI
> templates etc. .. but without being standards, who cares? WSDL 2.0's

It remains to be seen how useful these things will be, I certainly see
some benefit in URI templates, but WADL is questionable.   

In any case, I'd say most of those developers interested in web
applications, especially the newer AJAX, Mashups, Atom, RSS/RDF, etc. 
wave of apps, will be pretty interested in these things.

Besides, are the WS-* standards really the computing industry's endgame
for integration?   Premature standards do no one any good. 

> HTTP binding has (IMO) the 80-20 needed to make plain HTTP (I'm
> avoiding
> using the R-term) really describable and usable by programming tools
> in
> addition to humans but restafarians don't like it because it doesn't
> start by naming operations GET/POST/etc.. Oh well. 

The HTTP binding describes an interface so one can generate brittle
code that interacts with the web just like a statically typed SOAP
message exchange.  That doesn't sound like a win to me... 

Cheers
Stu


 
____________________________________________________________________________________
Want to start your own business?
Learn how on Yahoo! Small Business.
http://smallbusiness.yahoo.com/r-index

Reply via email to