Dave Remy Blog
Friday, July 18, 2003
Breakfast with Henry Thompson
I had breakfast this morning with Henry Thompson along with my colleagues from BEA, Cliff Schmidt, David Bau (architect of XmlBeans), and Kevin Krouse. Henry was in town (Bellevue, WA) for an XML Schema working group meeting at Microsoft. Henry is the editor of the Structures part of the XML Schema spec as well as the creator of the first publically available implementation of a schema validator (XSV).
Henry was great very approachable and full of background and information about XML Schema. Having David Bau and him together at the table talking about schema (as well as Cliff who was on the XML Schema working group as a Microsoft Rep) went beyond my understanding many times. Just in the short conversation one bug in XmlBeans validation was identified having to do with ALL and using a counting strategy with no-backtracking (which is what XmlBeans does). From what I gathered XmlBeans uses a counting strategy with no backtracking and there is a circumstance (which Henry sketched for us on the back of a coaster) where just using counting would not be able to validate. It's an edge case but I bet David will have a fix for that by tomorrow.
Another interesting subject was related to the Test collection. We (XmlBeans team at Bea) have been using this heavily to test XmlBeans however these tests are not normative or really backed by the W3C as being a full test suite for XML Schema performance. We found that many of the schemas in the test collection have not been updated to reflect errata so we had to update them (info on XML Beans schema compliance including zip of updated test collection) ourselves. I guess there was a licensing issue around "document" versus "software" licenses that is near resolution. There will be a subgroup of the Schema Working group focusing on testing and a process for submitting changes to the test collection has already been defined (great!). It would be great if Bea participated in this test group. From my perspective having a strong XML Schema compliance test suite would be the best thing that could happen to help XML Schema keep the momentum it has gained. There are just too many interpretations right now, as well as out of date implementations.
