W3C home > Mailing lists > Public > www-ws-arch@w3.org > August 2003

RE: Book chapter on Synch/Asynch

From: Cutler, Roger (RogerCutler) <RogerCutler@ChevronTexaco.com>
Date: Mon, 25 Aug 2003 12:09:35 -0500
Message-ID: <7FCB5A9F010AAE419A79A54B44F3718E01817F8D@bocnte2k3.boc.chevrontexaco.net>
To: doug@rds.com, "Francis McCabe" <fgm@fla.fujitsu.com>, "Sai Surya Kiran Evani" <evani@informatik.uni-freiburg.de>
Cc: www-ws-arch@w3.org

Thank you for the comments on time.  I think that those two situations
are ONLY abstractly the same if you have a particular abstract framework
in mind.  Time does, in fact, exist and there is plenty of experience in
handling the abstract concepts involved with time.  Having a background
in theoretical physics I am particularly aware of some of these, and the
classic paper by Lamport I keep citing indicates some other approaches
which are by no means inconsistent with the physics of time and which
have a firm and consistent theoretical foundation.

Looked at in terms of Web services, I firmly maintain that the ability
of a Web service to respond in a time frame that makes it reasonable to
use it synchronously (using the word in its common sense) is something
that characterizes the Web service itself, not any particular
implementation or usage.  Although that distinction is obviously not a
black and white thing (as stated), the endpoints are very clear and I
cannot imagine that it is beyond the wit of man to devise some sort of
objective descriptive feature of Web services that would be useful for
this purpose.  For example, and I do not want to hold this up as
anything more than a shot in the dark, one might have an advertized,
"Mean Observed Response Time" -- that is, a mean of the response times
(not including transit times for messages) observed in some interval of
actual usage of the service.  Please don't focus on this suggestion and
tell me everything that is wrong with it -- it's just intended as an
illustration of what a feeble wit can come up with in a very short
interval of time.  I'm sure you can do better.

-----Original Message-----
From: Doug Kaye [mailto:doug@rds.com] 
Sent: Monday, August 25, 2003 10:02 AM
To: 'Francis McCabe'; 'Sai Surya Kiran Evani'
Cc: www-ws-arch@w3.org
Subject: RE: Book chapter on Synch/Asynch

>Francis asked:
>     By event driven, I assume you mean an event driven
>	implementation design pattern?

Yes, that's what I meant.

>Kiran asked:
>	Would calling the message exchanges ordered/unordered instead of
>	sync/async be more clear?

Interesting question. I wonder if it's really all that simple. I also
wonder whether it's sufficient to ignore the realities of time. For
example, is there any difference between a request/response MEP that
spans a few seconds and one that spans a few days? One would certainly
expect a difference in implementation although they're abstractly the
same. It depends on what you're trying to define and for what purpose.

(Please forgive me if my responses are particularly naive. I haven't
been part of this discussion until now, and I understand there's likely
been much debate on this topic by others more qualified than me.)


Doug Kaye, CEO
RDS Strategies LLC
doug@rds.com, www.rds.com
v: 415.453.1400, f: 415.459.0103
Received on Monday, 25 August 2003 15:05:01 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:41:08 UTC