RE: Enlightenment via avoiding the T-word

["Simon St.Laurent" <simonstl@simonstl.com>]

On 26 Aug 2001 11:29:36 -0400, Champion, Mike wrote:
> Hmm, I'm not sure I understand your point.  Tim's saying (as I understand
> it), we CAN all agree that elements/attributes/namespaces describe "labels"
> for markup information.

Yes, though...

>  That seems to be simple enough to share, and is a
> solid foundation to build on, even if various groups go  off in their own
> direction with higher-lvel interpretations of these labels and structures.

This is where it starts to break down, largely because it isn't clear
that the interpretations are genuinely higher level.

Even without any expectation of a PSVI, I can say things like:
<my:person xmlns:my="http://example.com/my">
  <firstName>Chip</firstName>
  <lastName>Skillet</lastName>
</my:person>

And people seem to do that with or without the blessing of W3C XML
Schema.  (Looking through SOAP, it seems they've been doing it for a
long while, something I wish I'd noticed long ago.)

The labels are all present, but the relationship between the labels and
the structures (and between the labels) are ambiguous.  In a strong
sense, I just can't get an unambiguous ulabel for firstName and
lastName.  To deal with that, I have to write code which deals with much
more than ulabels.

I'd felt that XML at least provided a foundation on which we could all
agree what the labels and structures were, an agreement enforced by
basic syntactical structures specified in XML 1.0/Namespaces in XML.  I
don't think we have that any more.

Is that a real loss?  I'm not sure.  It does mean that I need to change
my marketing pitch, though.

> That is, I might build tools that understand XML syntax only as labels and
> structures (DOM Level 1 and 2, XPath, tools for validation modelled as
> structural constraints). Lots of people can build perfectly reasonable and
> funcitonal applications directly on this foundation. Somes believer in the
> utility of Schema t***s (oops, sorry Tim, wash my mouth out with SOAP) or
> the PSVI or the Query data model or the RDF/AF/whatever conception of what
> these labels *mean* can still layer their own conceptions on my more
> minimalistic conceptions.  Or they can build directly on the T-word
> conception with databinding or the PSVI ... that's their business, as long
> as they don't try to deprecate the "it's just a label, stupid" view of XML.

I think I'm saying that "it's just a label, stupid" doesn't work any
more, in that there are now diverse interpretations of labels and
ulabels. Labels which have different relations to structures (i.e.
content models) in different contexts are a bit of a challenge, but not
that hard to accomodate.  Labels which are interpreted differently based
on context and the nature of the reader and writer expectations are a
lot harder to accomodate.

> The problem -- as you noted in your XML.com article a couple of months ago
> -- is that the folks who want to drive "the T word" all the way down to the
> foundations of XML (or at least the PSVI) seem to be running the asylum
> these days, and this makes the rest of us inmates even crazier than we were
> to begin with.  This leads to the situation that you Simon and I bitch about
> continuously ... but Tim's offering a way out by clarifiying the distinction
> between labels, namespaced labels, and whatever theoretical superstructure
> people want to erect on top of those labels.

I don't think the label foundation can support such superstructures at
this point.  In a sense, we now have two unproven structures
(unqualified names and W3C XML Schema definitions) supporting each
other.  That can work - it's even an opportunity for creative openings -
but it leaves me feeling deeply unsettled.

> Simon recently proposed a fundamental differentiation between the "raw
> syntax" and "PSVI" conception of XML, and a proposal that the two groups go
> their own way (my memory is fuzzy .. forgive me if I've distorted the
> post!).  Tim is proposing, as I interpret it, a more highly refined and
> perhaps less divisive differentiation -- at the root, XML structures (be
> they syntax structures or parsed data models) are identified by "labels" and
> "ulabels".  The "PSVI" conception presumably accepts this as a foundation,
> but then builds a superstructure on top of it.
> 
> Can we live with that?

I don't think the foundation is solid enough at present for that to be a
safe approach.  I'd rather build beach huts on sand than skyscrapers, if
you know what I mean.

Simon St.Laurent
http://simonstl.com