RE: D-AR003.1

> -----Original Message-----
> From: Mark Baker [mailto:distobj@acm.org]
> Sent: Thursday, May 30, 2002 9:40 PM
> To: Champion, Mike
> Cc: www-ws-arch@w3.org
> Subject: Re: D-AR003.1
> 
> If we agree  that trying
> to hide application protocol semantics from Web services is one way of
> violating your requirement, then I think we should try and 
> capture that now (before you change your mind 8-).

I already have :~) Not really, but I just want to make clear that there is
an intrinsic appeal to the "hide the network" idea.  It would be nice if
applications didn't have to worry about checking to see if messages got
through, and re-trying idempotent operations "just to be sure" ... and
having to poll to see if an operation has completed yet, and the various
other housekeeping chores that RESTful applications must do.  It would be
nice if that While I might guess that this is the best way to build web
services over unreliable and insecure networks, I don't want the WSA to
*assume* that this is inevitable.  After all, Don Box might get his way and
see HTTP replaced with a reliable, secure, 2-way network infrastructure
<grin>. If that happens, a lot of what we agonize about will be moot -- the
Internet will look like a giant Intranet.   

Seriously, one can argue that the long-term trend is to move the
housekeeping details that assure reliability and security from the
responsibility of the application to the responsibility of the
infrastructure.  That's clearly what DBMS's offer in the storage arena, and
one can assume that sooner or latter networking infrastructures will take up
that burden as well.

I don't expect that anytime soon, but it could happen, and it *is* happening
as within-enterprise network infrastructures become standards-based
intranets.  So, I repeat my bottom line: the WSA needs a requirement to
cover both "behind the firewall" scenarios where network infrastructures can
be effectively hidden, and "outside the firewall" scenarios where the Right
Thing is to align the web service semantics closely with HTTP or SMTP/POP.

So,D-AR003.1  now says that the web services architecture "separates the
transport of data or means of access to Web Services from the Web Services
themselves."   

Mark suggests "supports Web service development where the semantics of
underlying
application protocols are exposed, but does not require that a Web service
use those semantics".

I'm thinking along the lines of "Allows the separation of the transport of
data from the web services themselves where appropriate, encourages the
alignment of data transport with the semantics of web services where
appropriate, and provides guidance for determining when one or the other is
appropriate."  Needs wordsmithing, fer sure! ... 

Anyway, Mark and I have been jabbering at each other for days about this ...
it really would be good to get other perspectives.

Received on Thursday, 30 May 2002 23:27:39 UTC