[BITS] Re: Does anyone besides me think that XML has no clothes?

I Find Karma (adam@cs.caltech.edu)
Wed, 25 Feb 1998 07:11:02 -0800

FoRKers: find this thread on dist-obj at


Kudos to Anders for delurking in such heroic fashion!

> [Anders] But of course XML has clothes. It's just not a solution to
> all problems in the world. Once and for all:


But, once we have eliminated syntactic incompatability (!), *and* we
have self-description (Win95 Registry may be global, but isn't
self-describing!), *then* we can at least envision proceeding to
semantic interoperability.

> [Ron] [XML is like the problem of business objects - JFOX and JTONE.]

But not by way of business object standards. Developing common XML
data models is precisely *not* the problem of shared behavior, in our
estimation -- that's proceeding from semantics back to syntax.

> [Anders] It's always difficult to get people to agree on anything. My
> guess is that it's not the syntax that has been the source of
> contention (given how ugly it is they can't have attached much
> importance to syntax in the first place). This must be exactly the
> kind of information that XML would be well-suited for.

The contention is behaviors. Absolutely right. However unlikely it may
seem to come up with a common DTD for price-tagging (currencies,
significant figures, quantities), it's exponentially more unlikely that
we'll come up with a common behavioral model for computing taxes,
discounts, frequent buyer points, and everything else different
organizations do *with* price tags.

In the other directions, standardizing the interchange of data (events,
IDL, and so on) is a prerequisite to semantic interoperability.

> [Mark] Remember my "ObjectWeb Context" post? In it, I described IDL
> as a DTD. It is more analogous to HTML than to XML. Namely, it's an
> instance of a structural declaration, not a self-describing
> declarative language.

Describing IDL as a DTD? This is only useful as a first step, really.
We agree with Anders here: "You can obviously encode IDL as XML - not
very interesting." He continued with some other pathways we liked in


> [several] [the airline example, repositories and dictionaries]

It is a complete red herring to talk about the interaction of Web
transport and delivery with dynamic DTDs.

First, URIs are names, not required behaviors to fetch. We can give you
a rock with an URL engraved in it, and that's it. No DNS lookup is
required, morally. Physically, look for 'stable' DTDs to have
UUID:names and other sorts of stable names so that the DTDs can be
compiled in or cached reliably.

> [Anne] XML parsers are likely to become a performance bottleneck.

We do not think there's much in current experience that says parsers are
necessarily a performance bottleneck. Disks and wires are much, much
slower than the efficient parsers allowed by the simplifications of XML.
Nevertheless, compact, efficient concrete representation *IS* important
for many sorts of applications. So rather than throw out the XML baby,
we posited some gedankenexperiments about what a "YML" might be:


It's not just the parsing, it's the tree layout. Viewing an XML stream
smeared out in space-time, it becomes more evident there may be call for
a "relocatable object format" that allows breadth- or depth- first
traversal of the XML tree.

[Actually, we'll just cross-post the YML post to d-o in my next message.]

> [Sundar] I've been somewhat bothered by all the hype associated with
> XML lately... I just can't seem to get how everything else is "just
> nails" as some of the XML hammer crowd would have you believe.

Talking about hype does not shed any more light on the subject. At
least on this list, let's stick to the facts here. XML will go places
that SGML never could:
- The parsers are simple.
- Optional features have been taken out.
- Namespaces are done in a more logical way.
- And so on.

> [Ian] SGML + hype + minor improvements
> and SGML has been around for quite a while now. If XML succeeds where
> SGML has failed, I suspect the "hype" bit will be the difference,
> rather than the "minor improvements".

XML represents some major improvements over SGML for the purposes of
networked document exchange -- mainly by pruning it down to size. On
the other hand, let's fully acknowledge that the books, conferences, and
lectures ARE a vital, valuable part of making XML real (Amazon now lists
14 XML titles and counting...). In game-theory models of standard
setting, such "sponsorship" can be absolutely vital, more so for
horizontal standards where lots of little players have a small personal
interest in building on top of it (a public good).

> [Sundar] inertia would be "against" such shared DTD's that everyone
> seems to be blithely assuming.

XML makes the easy things easy, and hard things possible. This is no
substitute for social agreement, but at least it prevents the computer
part from being the limiting factor.

> [Ron] I remain a dreamer at heart, and hope for a day of global
> semantic level interoperability where my object can query yours and do
> the handshaking thing at runtime, based upon common domain
> assumptions.

XML is NOT an object model. It's a *DATA* model. It's the opposite of
the OMG's conspicuous attempt to make interoperable objects, by taking a
data-centric view. Jokes about elephants aside, this is a significantly
different approach to traditional problems.

Not that data-centric electronic commerce hasn't been tried before: but
EDI wasn't self-describing (where would all those translating VANs make
their money?!). That's the difference: horizontal leverage from generic
tools that can readily manipulate extremely specific formats.

> [Ron] Here's my rule of thumb when reading things like Rohit/Adam's
> posts: - If it smells like magic, and you can't quite follow how they
> got from A to B, then it probably is magic, and you probably can't
> get from A to B :-).

This is uncalled for, Ron. This is a professional list, so let's avoid
the personal attacks. Try asking what step can't be followed. All we
saw was Doug Lea's {intriguing | witty | subtly relevant} link to
Elephant 2000. If you're having fun off-list discussions of our post,
please don't try us in absentia.

> [Ian] Personally, I really hope that XML succeeds! But technically,
> how many problems does XML solve that existing alternatives (SGML,
> CDIF, RTF, IDL, IIOP, Java serialization, etc.) don't?

The whole point is, XML solves ONE problem really well. It is not all
things to all people -- jes' one consistent toolkit for constructing new

Rohit and Adam


Do you realize there are only 5 degrees of separation (in Web space)
between the Macarena lyrics and this XML/HTML paper?
-- Dan Connolly