At the end of a rollicking, six-hour review of the "Bleeding Edge of XML", Elliotte Rusty Harold spent half an hour discussing the future of XML in very specific terms -- what would succeed, what would fail, and what would just squeak by.

After briefly noting the dangers of predictions, Harold proceeded to make them, starting with XInclude. This, he felt, would succeed as soon as parsers supported it. The ability to construct documents from parts of other documents was simply too useful to ignore and did not harm people who didn't use it.

He predicted a similarly rosy future for JDOM, "much to the consternation of the W3C." Noting that thousands of developers were already benefitting from its simplicity and ease of use, he called it the "triumph of worse is better" and went on to say that he thought Sun would endorse it through their Java Community Process. In spite of some reservations -- he cited experience implementing an XInclude processor in which JDOM lacked features found in DOM -- he still felt it fit the 90% case needed by most programmers.

Harold was less enthusiastic about the W3C's XML Schema Language, predicting "a partial success." He felt that developers needed schemas desperately, but also thought that the specification was too complex to be used widely. Furthermore, he suggested that schemas would be replaced within ten years, as soon as enough experience had been gathered to determine what was useful and what was not.

In a quick review of other specifications, he thought that XLink and XPointer would not succeed without a killer application, that XSLT, an outstanding success, would have little reason to change, and that XSL formatting objects would enjoy slow but steady growth, especially in the print community. He also predicted that both MathML and SVG would succeed, stating that both filled long-standing needs.

On the losing side were XHTML, JAXP, and the schema repositories. XHTML would fail partly due to complexity, lack of tool support, and poor documentation, but mostly because it provided no real benefits to Web page authors. JAXP, although useful now, would eventually succumb to functionality in SAX 2 and DOM Level 3. And the schema repositories (CommerceOne, UDDI, BizTalk, xml.org, and so on) would be victims of restrictive copyrights and little benefit to customers. In their place, he saw most schema development taking place in industry groups, which would self-publish, hopefully in the public domain or under a non-restrictive license like the W3C's.

Harold's last prediction was that browsers wouldn't support XML until 2002 and that, when support did come, Mozilla would regain market share from Internet Explorer. In spite of this delay, he did see browser support as an inevitability, spurred by fragmentation of the Web by non-PC devices such as cell phones and handhelds. These, he thought, would force the adoption of XML as a browser-independent format.

In closing, Harold was optimistic. Quoting Alan Kay, who said that, "The best way to predict the future is to invent it," he encouraged his audience to be an active part of that future.