Re: [All] changes to provenance

felix,
OK, I appreciate the changes are complex to take on board at this stage, 
making the logical nesting is getting very hard to read and they 
probably don't buy us much.

So we can stick to the trans/transRev version rather than using the 
explicit activity attribute.

Referring to the other changes made in:
http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Nov/0089.html

- made some wording updates on the description

>>just wording imrpovement - should keep these

- renamed the data category, actually just to provenance (the provRef
aspect amkes it slightly broader than just agents)

>>As you suggest

- removed the trans/transRev subdivision and introduced a
differnetiating flag 'activity'. Currently this just has values
translation and translationRevision, but we can quickly add to this as
suggested by marcis

>> As you suggest, we drop this

- made the use of pointer and non pointer rules mutually exclusive
(hopefullly examples will now tell us which we really need)

>> So shall we just avoid mixing them in the examples, rather than forcably prevent mixing pointer/nonpointer in the logical structure? I attach an XLIFF 1.2 pointer based mapping example we can use. I think this is in fact the only real use case for global pointers for this data category - the ones in the current examples are fairly artificial.

- allow stand off provenance records to be used in global _and_ local

>> the (probably only) use case for this is to allow provenance standoff to be used with attribute text. Actually, i think that if we change the inheritance rule from 'include element but exclude attributes' to 'include elements and attributes', the we can annotate attributes well enough with local markup - it will rarely be different from the surrounding element mark-up (unlike mtconfidence where every string will have a different score). So we could then drop non-pointer global rules altogether.

- I broadened the language for the provRef attribute, so it can still
point to a PROV record (specifically an entity one), but could also
point to other records. Seeing the work starting on XML change records
at TPAC and also seeing some convergence on a specific PEMT activity
logging format in XML between CNGL and the CASMECAT project - it seemed
prudent not to exclude these others (note this isn't bound to the
activity value like the other attributes)

 >> as stated, this is just being prudent, not a big issue.

Does this sound OK?

Dave

p.s. I'm checking through the other examples now also.



On 14/11/2012 13:11, Felix Sasaki wrote:
> Dave, all,
>
> since we only have two weeks time, I would propose to do only one 
> change to the definition of proveance: rename "translation provenacne" 
> to "provenace". All other changes (e.g. a different role for standoff, 
> see the other thread) would need more time, which we don't have. It 
> also will raise concerns from other implementors like Linguaserve, as 
> we have just seen.
>
> Thouhgts?
>
> Best,
>
> Felix
>
>
> -- 
> Felix Sasaki
> DFKI / W3C Fellow
>