On Apr 2, 2008, at 21:41, Dan Connolly wrote: > > On Wed, 2008-04-02 at 19:13 +0100, Geoffrey Sneddon wrote: >> On 2 Apr 2008, at 18:53, Dan Connolly wrote: >> >>> Please help me remember to follow up; i.e. find >>> test cases that distinguish the two designs and find out >>> what requirements, if any, motivate the differences. >> >> Sorry for jumping in on this like this (I saw this on -archive), but >> the answer is the normal one: there is a large amount of content that >> relies on the existing algorithms. There are known bugs in the >> algorithms, but I do have tests for what is currently there (at least >> for the numeric ones) at <http://hg.gsnedders.com/php-html-5-direct/file/tip/tests/numbersTest >>> (see the README file in the same folder for more info). > > Thanks for the pointer to test materials; if you can isolate > any tests where the XSD design would lead to different > results, I'd be much obliged. Part of the problem is that XSD datatypes don't have well-defined processing requirements against which to write test cases or to implement UAs. "[Definition:] error A violation of the rules of this specification; results are undefined. Conforming software ·may· detect and report an error and ·may· recover from it." http://www.w3.org/TR/xmlschema-2/ For example, it is defined what an XSD 'decimal' looks like, but parsing the value is not defined with well-defined behavior on error. HTML 5 specifies how to parse the values (with error recovery that is compatible with existing content). On the validity side, HTML 5 doesn't allow the space characters before or after the value but XSD does. Thus, the HTML 5 definitions implement both sides of Postel's Law. -- Henri Sivonen hsivonen@iki.fi http://hsivonen.iki.fi/Received on Wednesday, 2 April 2008 19:36:20 GMT
This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 7 November 2012 14:18:13 GMT