At Fri, 9 Mar 2007 08:02:17 -0800,
Brian Tingle wrote:
> Erik, thanks so much for posting this.
> Re: testing if the schema is overly permissive. Before I released the
> last schema, I extracted METS from all the registered profiles and
> tested them against the new schema. But I wonder if we shouldn't
> develop a test suite of bad METS to make sure specific things are not
> allowed by schemas (w3c or RNG or whatever)? Does anyone have any
> thoughts on this idea? Would it be better to have one bad METS per
> error we were testing for; or could it work with one *really bad* METS
> that had errors all over the place?
The one error that I had with the Relax NG schema was an empty fptr
showing as invalid. So, at least for Relax NG, it seems that the most
useful bad METS would illustrate these sort of structural issues,
rather than simple bad values or misspelled elements.
Of course, the easiest thing to do would be to gather all the bad METS
that people have found in the wild and use them as a test suite.