Ooops. That's definitely a bug. Investigating...
Tomaz Erjavec wrote:
> I've encoded a dependency analysis using the linkGrp element, and it
> currenty looks like this:
> <linkGrp domains="#F0028708.9.1 #F0028708.9.1">
> <link type="dol" targets="#F0028708.9.1.3 #F0028708.9.1.2"/>
> <link type="del" targets="#F0028708.9.1.6 #F0028708.9.1.4"/>
> <link type="tri" targets="#F0028708.9.1.6 #F0028708.9.1.5"/>
> So, the @domains identifies the sentence (as synt. links are always
> sentence internal, both domains are identical) and then each links point
> to the two tokens in a syntanctic dependency relation.
> To my surprise, the above is invalid, as @domains is defined as
> data.name, so can't start with a #. And indeed, looking at the Guidelines,
> the example given is
> <linkGrp type="imitation" domains="dunciad dunnotes">
> where dunciad and dunnotes are taken to be IDs of some elements.
> Now, I wonder why @domains aren't the same type as @target, i.e. 2–∞
> occurrences of data.pointer? Wouldn't this be better, as you can use the
> same mechanism for resolving the IDs as everywhere else? And what are
> you to do with the current specification if the domain is in another