Re: How namespace names might be used

Al Gilman wrote:

> Walter, it is not meet that semantics [even post-syntactic semantics]
> should be entirely private to the node.  This leaves out too much of the
> utility of communication using mutually understood encodings.

'Local to the node where instance markup is processed' (my phrase) and 'entirely
private to the node' (yours) seem to me to describe very different conditions. I
believe that the particular semantics elaborated through processing of instance
syntax on a particular occasion at a specific node are potentially unique. I
assume, however, that instance markup has arrived from another node and that the
very act of instantiating all, or as much of it as can be comprehended within the
epistemological boundaries of the receiving node is the very essence of useful
communication. The risk when two autonomous nodes both defer to a third party to
prescribe the semantics which they will (identically) attach to specified syntax
is that those semantics may be (differently, but equally) meaningless to each of
them. I begin from the premise that each node is the best authority for what it
can understand and for what it can do to elaborate a familiar understanding from
various syntactical expressions. I have developed and described the various
mechanics through which a node may use its history of communication as the basis
of determining how the idiosyncratic syntactic usage of other nodes should be
processed in order to elaborate the semantics which its own idiosyncratic
expectations require. Surely these semantics, which depend on the varying peculiar
syntax of other autonomous nodes, are not 'private', if by that you mean that the
problem elaborating them is cryptologic, rather than hermeneutic.

> This is why Simon, TimBL, Tim Bray and I, each in our own words, have all
> been talking up separation of concerns.  Whether you call it packaging,
> layering, or architecture, there needs to be a way for syntax and semantics
> to coexist in the the platform for effective communication without sending
> processing into a spiral of infinite regress.

I am talking about precisely the same sort of layering. I may simply articulate
the layers at a different place. I do believe that the minimum common denominator
of interoperability depends on original, or fundamental, messages understood on
their syntax.

> "Zero semantics" is a Solomonic judgement that this mother-claimant, at
> least, wants to reject immediately.

I am striving, through process, not for 'zero' semantics, but to elaborate from
simple syntax, considered in context, the most exuberant as well as the most
particularly detailed semantics possible. How have my words so often repeated been
so misunderstood? I apologize that I have communicated this so crucial point so
badly.

Best regards,

Walter Perry

Received on Tuesday, 13 June 2000 14:51:53 UTC