XML 2001 was held last week at Walt Disney Resort in Orlando, Florida. Meeting up with old acquaintances and put some faces to e-mail addresses I have long corresponded with on XML developer lists -- most notably xml-dev -- was great. It was a total roller coaster ride. A roller coaster ride for the mind, that is, as I did not even get near the theme parks, but all the twists and turns going on in my head left a real-enough feeling of nausea.

Formulating ones thoughts after an XML conference always takes time. XML conference attendees are such an eclectic bunch of techies, philanthropists, monopolists, nuclear physicists, lawyers, and musicians that the clashing worldviews invariably generate an exciting cauldron to swim in for a week.

Having done my fair share of swimming and emerging from the far side in one piece, I think my abiding memory of XML 2001 will be as the conference at which the data-heads and the doc-heads finally agreed to call it quits and go their separate ways. Sad but true. Speaking as a doc-head, I am hopelessly incapable of presenting both sides of the argument but here is how I see it.

The data people are steeped in two related technologies: Object- oriented programming and relational databases. Both the OO and RDB users see uses for XML as an information interchange and on-the-wire format. They talk of data binding, object serializations, and third normal forms. An entire family of tools is coming into existence to facilitate this view of XML.

The document people are a bit bemused by the goings on of the data people. Relational databases are very good for manipulating data that fits the relational model. Object serializations are good ways to persist objects. Adding pointy brackets is merely a point increment improvement over CSV (comma separated values) or, say, s-expression- based serialization notation. Data people clamor for XML tools and the doc heads think, "Gee. What's wrong with ODBC, JDBC, Visual Basic, JetForms, MySQL, etc."

The document people have the same problems as always: Editing tools are dogs when it comes to mixed content and attributes; rendering tools fall far short of what we need to even approximate decent Web presentation and/or paper typography. The ethos of separating content from presentation breaks down at the old chestnuts called tables, equations, and graphics -- just as it always has.

Document people see the world in terms of XML encoded information flowing through systems, perhaps undergoing transformations and validations at various stages along the way. Data people see rigid XML structures flowing over the wire between well-defined end-points that encode all the really interesting stuff in "business logic" at the end- points.

The data people point out that, with a little effort, everything can be reduced to chucks of stuff that can be molded into a relational worldview. They are right. Document people point out that relational databases are horribly restrictive and do not naturally fit the way the world of knowledge seems to work. They are right too. The RDF and Topic Maps people point out that both view are right and wrong simultaneously. At a higher level of abstraction, its all-just relationships between atomic units of data, some of which are also relationships.

Amidst all my cloudy thoughts and the conference-generated whirling in my head, I believe we are at the birth of the first major fork on the XML road. The worldview of the data people and the document people seem too different to be reconciled any time soon. The irony here is that unless I'm crazy, the two forks in the road will re-merge in part two of the Web Services revolution.

Sigh. Maybe it is necessary for both communities to go their separate ways in order to find out how much they really like each other and have in common. Anyway, we will see. In the mean time, hear this: If you have not been to an XML conference, then you should try it. A real roller coaster ride that does not involve a theme park, but generates the best thrills, and scares, in your head.