Re: inclusion

I find this a convincing practical argument for Simon's earlier conclusion that
the 'lower', 'syntactic' level consists of XML 1.0 + Namespaces (literal
interpretation, as written), period. Everything else, Infoset included,
introduces idiosyncratic semantic intent which can be (idiosyncratically)
worked out within each respective spec (and, with further idiosyncracies, in
its implementations), but which is unlikely to be entirely consistent with the
idiosyncrasies of other specs. I am not troubled by this. In fact, it
illustrates anew that the interoperability of XML is inherent in the syntactic
layer:  every use or re-use of XML content is predicated on a new parse. On top
of that parse, processing idiosyncratic to each spec, to each implementation of
that spec, to each processing node, and presumably to each instance of
processing will yield unique results, uniquely appropriate to the
circumstances. This is a reasonable--and often necessary--outcome of the
autonomy of the node. Interoperability between nodes and between instances of
processing over time is unimpaired by this predictable divergence between
pipelines of process.

I now feel that this discussion has answered not only the original
question--which I doubt that many of us ever expected it to--but has also
re-affirmed the soundness of XML 1.0 as a syntactic spec and taught a
hard-knock lesson about the nature of more-semantically-loaded specs which can
be written on top of it. I am sure that others here will draw different
conclusions from mine and that even those of us who come to something close to
my present conclusions will need to re-learn them some months down the line.
Nonetheless, I am satisfied in ways that I never expected to be with how
inevitable from the fundamental nature of XML was the conclusion I now believe
that this discussion has reached.

Respectfully,

Walter Perry


"Simon St.Laurent" wrote:

> I keep coming up with ugly cases, and the complexity levels involved in
> absolutizing relative URIs appear to be dazzling, though I'm not sure this
> latest case is actually good for any side of the argument.

Received on Thursday, 25 May 2000 03:29:44 UTC