Lists Home |
Date Index |
> People who use such tools deserve to lose: if you can pass an element
> name of "@@@@" (which is no more valid 1.1 than it is 1.0) and get a
> purported XML 1.0 document written, then you might just as well go
> back to printf (or equivalent).
Umm, I disagree. There are many situations where this occurs in a pipeline
process-- as the output of another SAX stream of events. Think XML
formatters based on SAX. Currently such standalone filters have an
expectation of 1.0 names (as guaranteed by the SAX process)... with such a
change that expectation is betrayed. Such filters would have never received
"@@@@" from a SAX event stream to begin with...
> Remember, a 1.1 parser MUST accept 1.0 documents, and a document
> generator SHOULD generate 1.0 unless 1.1 is required by the content.
> Thus saith the 1.1 Rec.
Right... I am referring to non XML 1.1 filters. e.g.
XML 1.0 goes through SAX pipeline gets to XML 1.0 output filter
The content "foobar" gets parsed. SAX processor chokes, it never gets
to the output filter.
XML 1.1 goes through SAX pipeline that accepts 1.1 gets to XML 1.0 output
The content "foobar" gets parsed. SAX processor accepts it and passes
it on-- it is output through the 1.0 filter mistakenly.
Granted this is just a versioning issue-- the moral of which is: don't use
old filters with newer SAX parsers which support 1.1 because they won't
work. It makes sense... but I also don't think that the filters as written
were inherently evil and that their users need to lose-- they just need to
upgrade everything when upgrading.