its:idValue and xml:id

Hi,

I was working on some XSLT that adds ITS markup to an XML and was wondering
on how to explicitly add ID's that other generic XSLT can pick up to
process (e.g. a generic XML+ITS to XLIFF conversion step).

In my case I'm not using the full power of ITS rules but rather use it to
make localization information explicit available on the elements. The XSLT
contains the rules in the form of templates. It will add @its:translate to
the translatable elements.

Now when I look at how to provide an ID that a next step can pick up I
believe the spec wants me to add an @xml:id (8.14.2)

Because I like to separate the processing that does the ITS annotation from
the XLIFF preparation as much as possible this @xml:id in my opinion breaks
this separation. A generic post processing step cannot simply throw out all
ITS namespaced stuff to get back the original XML. In case of added xml:id
attributes it has no way of knowing if this wasn't already present in the
original or if it was added by the ITS annotation step.

I always thought that this was valid use case of ITS. That you could have
an XML+ITS document where all ITS information is made explicit on the
elements itself and no ITS related processing (like rule interpretation via
its:idValue) was needed to use it. Hence I rather expected to see an
@its:id to provide localization IDs.

I haven't thought this through for other ITS datacategories and rules but
for simple stuff such as @its:translate, @its:locNote etc. I can work like
that but not for IDs.

Any thoughts?

-- 
--Marc