Cart
[Home] [By Thread] [By Date] [Recent Entries]
From: Simon St.Laurent <simonstl@s...> >At 09:49 AM 5/2/01 -0600, Arnold, Curt wrote: >>Congrats to all on reaching recommendation. Now on to fully >>implementing and testing the thing. > >Yikes! I thought the the CR phase was supposed to be about implementing >and testing. CR phase trials the implementability of the spec. There is a chart somewhere showing, feature by feature, which features had been successfully implemented. When there were many (say, 3 or more) different successful implementations of a feature, the WG deemed that feature to be "implementable" and considered the feedback from developers. In some cases implementation was slow in coming: key/keyref/unique springing to mind. We can take that to indicate that either the CR spec for key/keyref/unique was too hard, or it was deficient, or it was not a "must-have" feature in the minds of implementors, or that implementors were having difficulty getting that far within the 3 month CR period. Consequently, key/keyref/unique were simplified for XS 1.0. I think there were a couple of other features in the same boat--perhaps dates, though both also had new information coming along even at the last moment, which is why the spec took fairly minimalist and revisable approaches to both dates (just provide Gregorian with no hierarchy) and key/keyref/unique (very simple Xpath subsets.) On the other hand, the test collection allows testing the implementations of the spec, so it is quite different in purpose from CR. It is _also_ (but not primarily) a systematic way to get feedback on the areas of the schema specs which confuse implementers, both for the technical content and the wording; but the primary purpose should be to allow one implementation to read a schema, create the components, assess a document, do a Richard's Dump of the PSVI, and verify that PSVI against the expected one. Cheers Rick Jelliffe
|
