RE: Proposed replacement text for Section 1.6

 

> -----Original Message-----
> From: He, Hao [mailto:Hao.He@thomson.com.au] 
> Sent: Sunday, January 11, 2004 11:23 PM
> To: 'Champion, Mike'; 'www-ws-arch@w3.org '
> Subject: RE: Proposed replacement text for Section 1.6
> 
 
> 1. The architectural goal of SOA (and WS in general) is to 
> "achieve loose-coupling between interacting software agents 
> in order to preserve the benefits of reusability, 
> extensibility and simplicity."

I don't think loose coupling is a goal; it's the result of good application
of SOA, Web services, etc.  My "goal" is to build a system that works
simply, reliably, and accomodates change.  "Loose coupling" in a software
engineering term to describe an architectural arrangement that is purported
to help achieve these goals.  I realize that these are fine distinctions and
don't want to argue about them, and would accept your proposal if others
like it.

> 
> 2. Two main architectural constraints of SOA: 1) A small set 
> of simple and ubiquitous interfaces to all participating 
> software agents. 2) Descriptive messages delivered through 
> the interfaces.  

OK.  I do see these as the key factors that distinguish SOAs from other
similar architectural patterns.  Calling them "constraints" is fine with me.
> 
> I, personally, would also add extensibility as part of the 
> constraints but Dave O would argue it is just a best practise 
> (however, he believes that extensibility is important and has 
> written a number of articles on it). 

I'm with Dave Orchard on this.  Extensibility is important, but it's a
consequence of doing things according to SOA constraints and best practices,
not a constraint in and of itself.
> 
> As to the relationships among the terms  "distributed 
> system", "service oriented architecture," and  "web service", 
> I believe there are just two main kinds, those based on OO 
> and those based on SOA. The confusion comes when one tries to 
> do "distributed objects" using Web services. 

Well (not to re-stock the ur-troutpond) you CAN do distributed objects with
web services technologies, and I more or less insist that we acknowledge
that.  To deny it would be to say that the first generation of Web services
tools is useless.  I think even the most fervent RESTifarians would admit
that they have their uses in secure, reliable, fast domains (such as a
corporate intranet). The issue has always been whether the the distributed
object approach scales to the cross-enterprise or Internet level, and if it
does whether CORBA is a better platform to build on than the WS-* stuff.  

Roger seems to disagree, and if we were trying to *prescribe* what the
industry should do rather than *describe* what it actually does, I wouldn't
have a problem with making a stronger statement about the folly of using WS
technologies to implement distributed objects outside the firewall.  But I
don't think Web services as we define them, with SOAP, WSDL, and XML as
necessary attributes, implies either an SOA or distributed object
archictural pattern.