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

Re: Snapshot of Web Services Glossary on Response types

From: Walden Mathews <waldenm@optonline.net>
Date: Tue, 25 Feb 2003 11:57:51 -0500
To: Dale Moberg <dmoberg@cyclonecommerce.com>, Assaf Arkin <arkin@intalio.com>, www-ws-arch@w3.org
Message-id: <003601c2dcef$09235480$1702a8c0@WorkGroup>

I have two questions about this:

(1) Web Services are protocol-independent, which makes me
fear that sometimes they may be connectionless, and then you'd need
to tweak your definition, wouldn't you?

(2) You're appealing to common usage in a narrow field as your
basis for deciding the definitions.  That's a reasonable approach, in
my view, but the glossary introduction kind of forbids it.  What's
your take on that?

Taking hints from the glossary intro, I'd suggest adopting a def'n
that is linguistically and historically consistent (soft sciences at best,
but what the heck), although it may not cover all the intended
meanings.  Then adopt other terms to regain coverage.  In your case,
"connection-bound" or something might work.

As Mike Champion says (and I hate agreeing with him on this ;),
there's not going to be any perfect solution.

Walden

----- Original Message -----
From: "Dale Moberg" <dmoberg@cyclonecommerce.com>
To: "Walden Mathews" <waldenm@optonline.net>; "Assaf Arkin"
<arkin@intalio.com>; <www-ws-arch@w3.org>
Sent: Tuesday, February 25, 2003 11:17 AM
Subject: RE: Snapshot of Web Services Glossary on Response types


>
> I normally interpret people to mean the following when they talk about
> the difference between a synchronous or asynchronous _response_. There
> are lots of other meanings to these terms as applied to other entities.
> I would not disagree with people lamenting the terminology selected, but
> it is a done deal at this point, IMO.
>
>
> A web service response is said to be synchronous iff it is returned
> using the same network connection used in sending the request to which
> it is a response.
>
> [This implies that only one URL would be needed for request-response MEP
> when the response is synchronous and uses a transfer protocol that has
> URLs. It also implies that the request and response occur within the
> interval of time that the network connection exists. Also, there is
> overhead in setting up only one TCP connection when TCP is used, and
> since we are talking about IP _connections_ that will be almost always
> the case.]
>
> A web service response is said to asynchronous iff it is returned using
> a network connection that is distinct from that used for sending the
> request to which it is a response.
>
> [Implications: Two URLs are needed to configure a request-response MEP
> when the response is synchronous. The response connection may occur
> while the request connection is still open or after it is closed. For
> TCP-based transfer protocols, two connection setups will be needed.]
>
Received on Tuesday, 25 February 2003 12:00:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:15 GMT