W3C home > Mailing lists > Public > www-ws-desc@w3.org > September 2006

Re: Proposal for WS-Types :: Web Services Description Specification

From: Amelia A Lewis <alewis@tibco.com>
Date: Tue, 26 Sep 2006 11:09:43 -0400
To: "Tyrell Perera" <tyrell.perera@gmail.com>
Cc: www-ws-desc@w3.org
Message-Id: <20060926110943.37d18117.alewis@tibco.com>

Well, hmmm.

Disclaimer: I do not speak for the working group (although I am a participant, representing my firm).  The following is not, however, an official position of my firm; it's just my opinion, based on experience.

RPC is not revolutionary.  Or new.  Object remoting is RPC in drag.  RPC is broken, in my opinion and experience; closing your eyes tightly and repeating "there is no network" does not lead to robust and reliable communications over the network.

There's a network there.  The network is not a computer.  Computer-internal addressing formats are not appropriate across a network, and the temptation to build in "pointers" inevitably leads these APIs into darkness and chaos.  There is no stack; there is no return value to pop off the missing stack.

"SOAP" is no longer an acronym, but when it was, it was the "Simple Object Access Protocol."  Yup, it was a revolutionary object-oriented protocol (with looser coupling than RMI or Corba or DCOM).  It is worth noting that the trend in web services has been notably *away* from "rpc style" to "document style"--toward messaging, toward looser coupling, toward asynchronicity.

I note that the last paragraph of page 3 of this proposal acknowledges that there is a problem: the proposed WS-Types-generated services are incompletely compatible with existing services.  That "object expiration" is tied to this raises awkward questions about distributed garbage collection.  That there is no mention of WS-Addressing in the discussion of addressing syntax is also noteworthy.  What about security?  Setting policies?  Is anything other than request/response in client/server mode supported?  For that matter--where's the schema, DTD, RNG schema, or pseudo-schema for the two XML dialects which are demonstrated here solely by example?

I'd certainly oppose any adoption by the WSD WG; I think that our group needs to finish getting the basic description of sending web service messages over a network done.  Past CR, into PR and then Rec, and we can all have a final blowout at the Last WSD Face-to-Face ever and generate embarrassing stories to tell about one another.

I would also recommend that the authors of this spec investigate WS-Addressing (for dynamic addressing of services), the WS-Resource framework (at OASIS) for something rather object-like that is more nearly native to the network, and build anything on top of other specifications that can handle some of the hard issues (where "hard" means "unsolvable in the general case").  For that matter, investigating WSDL's extension mechanisms and something like WS-Policy, to provide annotations to WSDL rather than a completely new XML dialect for description of messages (oh, all right, then, "messages masquerading as objects") over a network.  Doing so would lower the barrier to entry (if it looked interesting, the people who still think that RPC is a good idea would prolly give it a whirl, after all ... but not if they have to give up existing stuff and compatibility with WS-*).

Uh.  I'd call that my two cents, but I s'pose it's a either a nickel (measured by size) or nothing (measured by encouraging statements).  Oh, well.

Amy!
On Tue, 26 Sep 2006 15:58:25 +0530
"Tyrell Perera" <tyrell.perera@gmail.com> wrote:

>Hi,
>
>We have been working on a Object Oriented Web Services Description
>Specification.
>
>"Here we suggest a new revolutionary web service specification that
>will totally change how people use web services or how people develop
>web service client applications. Currently web services do not support
>OOP style programming. Today, the objects returned from the web
>services only have properties. They do not have any methods.
>
>This new technology will enable web services to return objects with
>both properties and methods. When these methods are called they will
>call some other web services. Since the service URL of the web
>services can be assigned dynamically, we can use them to develop
>service brokers that work seamlessly."
>
>We are attaching a preliminary draft proposal to this mail. Please let
>us know your thoughts on this and necessary steps to follow, in order
>to bring this to community evaluation.
>
>Thnaks and Regards,
>
>Tyrell Perera and Ruwan Wijesinghe
>


-- 
Amelia A. Lewis
Senior Architect
TIBCO/Extensibility, Inc.
alewis@tibco.com
Received on Tuesday, 26 September 2006 15:09:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:41 GMT