W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

Re: Syntax and semantics

From: <keshlam@us.ibm.com>
Date: Tue, 23 May 2000 16:25:39 -0400
To: "Chris Angus" <Chris.Angus@btinternet.com>
cc: xml-uri@w3.org
Message-ID: <852568E8.0070340E.00@D51MTA03.pok.ibm.com>
>I would suggest that it is not the semantics that must be
>carried by the message but rather sufficient means to allow an intended
>recipient of the message to identify the same mapping between the message
>and the meaning of the message as that used by the sender.

Bingo. Examples have been previously cited where the same message is
interpreted under multiple semantics as it flows from point to point
through a business process. The fact that I have a mailing address does not
always mean I'm using it to cut a shipping order; sometimes I'm using it to
do traveling-salesman routing for a service tech, or to do statistical
analysis of what's selling where.

Semantics are _LAYERED_. The first layer is reliably recognizing the data.
Only after that has been achieved can we map that recognition to the
operation(s) we want to perform.

Namespaces operate at the recognition layer. They're an extension of the
element and attribute names. They're a useful tool toward achieving the
higher-level semantic bindings, but are not a complete solution by itself,
nor intended to be one.

There may or may not be a way to support TimBL's vision of "In a
distributed system, the semantics must be carried by the message". But as
has been shown, the choice of syntax  for the namespace name does not
particularly affect this effort. Explicitly binding the namespace to
additional data achieves precisely the same result as attempting an
implicit binding via the namespace name, and if anything is _more_ flexible
and powerful.

I still fail to see a justification for stretching namespaces out of their
intended shape to achieve a goal better reached in other ways. Using the
namespace name would have been a cute shortcut, _IF_ that choice had
cooperated cleanly with the needs of namespaces themselves. It's pretty
clear to me that it doesn't, and that it wouldn't be the best solution in
any case.

I'm afraid that  the debate may have broadened the gap rather than closing
it. I think there's a larger consensus that the literal solution is the
best we've got, but I think that those of use who were involved in the last
go-round have generally only confirmed that, yes, we really did consider
all these issues  and the poll answers did say exactly what we wanted to
say -- that none of the answers were delightful but that
Literal-and-discourage was the best answer available to us.

We may be closer to a consensus.  I'm not sure we're closer to an

Joe Kesselman  / IBM Research
Received on Tuesday, 23 May 2000 16:25:59 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 23:13:58 UTC