[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[dist-obj] DOP & web-services & n-tier


I think Mark mentioned this article in one of the earlier threads.

I've heard or read Bosworth's take on these themes in the past, and I
guess I'm not yet sold on their significance.

- On loose coupling and SOAP (vs. Dist Objects), I don't follow.  The
protocol and the wire format are the key, he says, and then he waves
his hands and talks about sending XML documents.  Can't see how SOAP
does this better than IIOP, or why the stuff about contract and
implementation is any different then what the OO world has been saying
for eons?  (Besides the state-maching angle which comes later... we've
crushed this territory here recently, obviously.)

- I've also had some trouble with this "loose-coupling and the web"
line of thinking (even though its intriguing).  Sure, http has a real
simple interface, the browser doesn't break when you muck with your
web site.  But that's cause all it does is display information; the
user, who's a human, does all the necessary adaptation to allow the
loose coupling.  How does this lesson apply when a program is the
ultimate consumer of the content?

- On the coarse-grained theme (i.e., distributed services should
support coarse-grained interaction), all I can think is that when
you're conditioned to thinking of objects/dist-objects as COM, this
comes as a revelation.  COM's bias was always fine-grained interaction
because COM grew from the desktop.  CORBA's bias has always been
coarse-grained since it was always about distribution.  

Asynchrony, and the need to define a state-machine style contract for
interacting with a server, is fine, and legitimate, and useful for
non-trivial services.  And if you define these things (formally or
not) for a Soap/Corba/whatever-based service, then what?  My guess is
a human programmer reads it, decides if this is a service she wants,
and makes sure she programs her client to interact in the prescribed
way (I guess I can imagine semi-automation of the client programming
process).  Did SOAP help?  Could a client program automatically
discover this service (without knowing the "schema" in advance) and
interact with it in the right way?  Is that a piece of this vision?
I'm not seeing it.

IMHO, web services happens when a critical mass of useful services are 
built.  All this hype over enabling infrastructure seems misplaced.  

pb


Eugene Kuznetsov writes:
 > If you haven't seen this already, this an interview of Adam Bosworth. It's
 > one of the more high-bandwidth discussions of how Data Oriented Programming,
 > web services and "loosely-coupled" inter-relate.
 > 
 > http://www.xmlmag.com/upload/free/features/xml/2001/06jun01/ab0103/ab0103.as
 > p
 > 
 > 
 > 
 > \\ Eugene Kuznetsov
 > \\ eugene@datapower.com
 > \\ DataPower Technology, Inc.
 > 
 > 
 > 
 > ==========================================================================
 > To manage your subscription, mailto:dist-obj-help@distributedcoalition.org 
 > Archives, FAQ, etc.     http://www.distributedcoalition.org/mailing_lists/

==========================================================================
To manage your subscription, mailto:dist-obj-help@distributedcoalition.org 
Archives, FAQ, etc.     http://www.distributedcoalition.org/mailing_lists/