Bill de hÓra

Hehe, the only contact I've ever had with Elliote was my stating that an implementation idea he was looking for feedback on would have issues with 1.1 and his saying the above. John Cowan was on the list where this took place but wasn't drawn on it.

In fairness to Elliote 1.1 allows some weird stuff. In fairness to 1.1 it seems to be intended to coexist with 1.0 rather than replace it, some of the decisions (in particular the hardest to justify - allowing for the encoding of C0 controls) came because some people demonstrated that they did need it.

That should satisfy those who belief that practical concerns are what should move development of spec's forwards. Those who prefer to think more in terms of theoretical "purity" and like consistency amongst a suite of specs will be glad to see Charmod referenced, albeit at the level of SHOULDs and MAYs.

The biggest change is the productions for Char, NameChar and the (renamed) NameStartChar. Some possibilities prohibited for names in 1.0 are pretty hard to justify. That said I think 1.1 is too lax, and should have reflected the property values of assigned characters in the astral planes better and been more explicit in how one should use characters that aren't currently assigned when they are assigned.

So, XML1.1 is an abomination, but with parts that are great and which we'll hopefully see survive into any 2.0 (some should be considered best practice with 1.0). You might need it, but you shouldn't use it if you don't.

Any version of XML beyond 1.0 is an abomination.

I agree with that this should be a short book and better should not have been done. See http://sqljunkies.com/WebLog/mrys/archive/2004/02/05/972.aspx for more.