- From: Savas Parastatidis <Savas.Parastatidis@newcastle.ac.uk>
- Date: Tue, 30 Sep 2003 16:15:23 +0100
- To: "Tom Jordahl" <tomj@macromedia.com>
- Cc: <www-ws-desc@w3.org>
[snip] > > I find myself agreeing with Tom (in spite of his liking of *Underworld* > :)), but then, I've been easily swayed on this topic. I wonder if the > folks who find "operation" too suggestive of objectness find attributes > similarly misleading. > Saying that "attributes" are just "suggestive of objectness" is an understatement :-) I have expressed my worries in the past that additional semantics are added to web services that simply don't exist. What does it mean for a web service to have attributes? A web service only knows about message exchanges. Whether you call it an "operation" or an "attribute" it's still a message. So why have both? In object-oriented systems attributes have meaning, have semantics. What's their meaning in Web Services? I don't believe in the "Grid requires them" statement. But even if that's the case, a separate specification could be defined. However, I am confident that it is possible to build Grid applications using Web Services without the need for attributes (or any of the other Grid-specific semantics given to services for that matter). What's next for WSDL? Private/protected/public keywords? Or, what other semantics should we add to web services? Ability to garbage collect them (my favourite), mobility, transience, transactionality, secure, composable, etc? (Before I am misunderstood... I am not suggesting that such additional features should not be implemented... well, for some at least. Instead, as it is the case with the factorised nature of WSA, a new functionality should be introduced separately without overloading the semantics of a Web Service). Best regards, .savas.
Received on Tuesday, 30 September 2003 11:15:44 UTC