RE: Protocol independence

Ugh - feel free to move the www-ws-arch thread over here, Mike.

> > Suppose we had a tuple space based system, such as one 
> based on Linda
> > or JavaSpaces.  Would it still be the case that a Web 
> services approach
> > to using this system would be to tunnel "getStockQuote" 
> messages through
> > write() or take() or other tuple space operations?
> 
> I would think that a consumer could write() an object that was a 
> representation of the Infoset defining a getStockQuote SOAP 
> message, and 
> then the service could take() the message, then write() an object 
> representing the Infoset of the response.  You'd need a way of 
> "serializing" the Infoset as a Javaspaces object (that's 
> sortof a mind- 
> bending concept) but I don't see why it couldn't be done.

Sure it could be done, but that wasn't my question.

I described a system which is already perfectly capable of
retrieving information like stock quotes, using read() or rd()
(and without any humans involved even!! 8-).  My question was,
effectively, whether a Web services approach necessitated
treating my tuple based system as a SOA?  And if so, why is that
a good thing when tuple spaced based systems are already
complete (and have been proven equivalent with message passing,
blah blah).

> > I'm just trying to understand the motivation for this view.  Thanks.
> 
> Uhh, because there are a lot of protocols out there, many of 
> which do the 
> job they were designed to do better than HTTP can?  

Who said anything about HTTP?  8-)

MB

Received on Thursday, 3 April 2003 15:18:40 UTC