RE: [xml-dev] Ten Years Later - XML 1.0 Fifth Edition?

[Michael Champion <mc@xegesis.org>]

I agree that this scenario is likely to occur.  I'm assuming, however, that the pain will be felt by the producer of the 1.0.5 document more than the consuming application.  People who need the wider set of name characters will have the burden of determining that downstream applications can actually support them; application developers won't in general have to worry about the possibility that someone will send them a document that will be rejected, any more than they do now about the possibility of getting an ill-formed document that it might be good business to accept.  (Businesses today presumably don't reject million-dollar purchase orders just because an angle bracket is missing somewhere, likewise they won't reject one with a fragment of markup in Cherokee in some open content field).

 

I do agree that "someone who deploys such an application is entitled to prevent failures happening by configuring it to use a parser that enforces the rules his application is relying on". Assuming this change goes through, the current thinking at Microsoft is that even if W3C treats it as a bug fix rather than version delta to the spec, we will NOT treat it as a mere bug fix in the XML libraries. That is, people won't run Windows Update and have the behavior magically change, it will be implemented in a major release of Windows / .NET / etc.. The previous version of the library will continue to be distributed for some time, and could be run side by side with the one that supports 1.0.5 so that developers can configure apps to enforce the appropriate rules.

 

[note this this is just my report of current thinking, don't take it as a commitment about future products, which I have no power to make!]

 

Would that "old-new libraries side by side" approach work in the Java or *ix/OSX worlds?

Michael Kay <mike@saxonica.com> wrote:

 

 The interesting case is an application that (a) handles all XML 1.0 names correctly, (b) can't handle all XML 1.0.5 names, and (c) actually starts receiving documents containing XML 1.0.5 names. I suspect that won't happen very often. But it doesn't alter the fact that someone who deploys such an application is entitled to prevent failures happening by configuring it to use a parser that enforces the rules his application is relying on, and that's going to be a lot easier to achieve if different versions of the specification are properly numbered.