Re: ECMA harness

I'm still not quite clear what the motivation is.  If the motivation is to be able to run the suite on an implementation whose ECMAScript does not support try/catch clauses, then that places the requirement that the test framework not use exceptions internally either.

I wouldn't go through that effort for obsolete browsers, while it might be interesting to see how well IE 3 conformed to the HTML DOM, the benefit of the test suite is to improve the conformance of existing and forthcoming implementations.

De-selecting is already in place.  If you declare that a test requires entity preservation and the implementation expands entities, then the test will be skipped.  We might need to create some new implementationAttributes, but I'm not sure what would be needed.

The best user experience is to run all the tests for a particular spec and have no test failures or at least only a few.  The best way to achieve that is to improve the conformance of the implementations.  Dividing the test suites up, so the user only sees a few issues per section doesn't make things substantially better.