Re: Error recovery

On Wed, Dec 5, 2012 at 5:34 PM, David Carlisle <davidc@nag.co.uk> wrote:

> opening
> up the data model has repercussions that are not impossible or even
> necessarily hard to specify but don't just automatically happen without
> specification as there are choices that need to be made.


I agree you can find cases (particularly in XSD-based specifications) that
would become unspecified if the data model is relaxed.  In practice, I
don't think these are a big deal.  XML 1.0 5th edition introduces very
similar issues (when used with specs based on pre-5th edition XML 1.0), and
the world has not collapsed.  HTML lives with a very similar situation: you
can have HTML DOMs that cannot be serialized using the HTML syntax.

I think the most helpful way to deal with these issues is to provide an
optional fixup stage that can be used in the rare situations where a
relaxed data model causes practical problems.

James