Re: REST wrap-up (was Re: Web Services Architecture Document

One last time into the trout pond, for the record:

> a) you believe that pervasive agreement on a form of solution is a
> sufficient criteria for success, no matter what form that solution
> takes?

Metcalfe's Law implies this.  All the hoary examples apply, like 
Betamax: it doesn't matter how good your VCR is if you can't rent tapes 
that play on it.  If all my business partners use a particular style of 
interface, that's what I have to use.   Even if (which I don't grant) 
you were right about the virtues of REST and the vices of SOAP/WSDL, 
asking an industry consortium to declare that common practice in an 
industry is simply wrong and that a niche approach should be declared 
the One True one is simply pointless.  W3C is an industry consortium, 
not sortof an Acadamie Francaise that enshrines  an ideal, holding  
back the tides of corruption by crass reality.

>
> b) your understanding of the architecture of very large scale systems
> suggests that a large degree of visibility[1] is not necessary

"Visibility" is precisely why XML is so pervasive these days.   XML, 
not HTTP is the secret sauce of Web services as far as I'm concerned, 
for pretty much the reasons Fielding notes in the paragraph you cite.

>
> c) you believe that Web services have sufficient visibility

The whole point of SOAP IMHO is the header/extension model; the focus 
of the WS-* is on standardizing headers and processing models for 
common business challenges (authentication, encryption, message 
correlation, etc.).  Those are the things Fielding talks about as 
virtues of visibility, so yes -- Web services have more or less 
sufficient visibility.  The major architectural challenge that WSA 
identified in the WS space, however,  is the lack of support for 
intermediaries in WSDL, i.e., one mechanism that exploits visibility is 
not exposed declaratively.

>
> e) some other reason(s)?

Roger said

 > it's even possible that it would have worked better and that it
 > would have been better had it been done that way.

Sure, it's *possible* that things would work better had the world 
jumped on the REST bandwagon rather than the SOAP/WSDL bandwagon, but 
two years of this discussion has yielded essentially ZERO evidence for 
this.  The one example being cited is Amazon.com's web service 
interface; the XML over HTTP interface is more popular than the SOAP 
interface.  To the best of my knowledge, this is completely consistent 
with my conclusions (which may not be expressed in the WSA Note):  
RESTful interfaces are probably best for read-only, non-confidential 
services over the public Internet (so people use the simple HTTP 
interface for querying for books).  Once they start needing to do 
business, however, they need more features that the conventional web 
services tools provide out of the box, so the B2B users of Amazon tend 
to use the SOAP interface.  This *could* be done with just XML and 
HTTP, and some extremely skilled XML/Web developers such as the 
Propylon folks seem to be having good success with it, but there's 
little publicly available evidence that this means that "it would have 
been better had it been done that way" for the industry as a whole.

Received on Friday, 6 February 2004 09:18:50 UTC