RE: issue-51 too many global rules

It seems that having global rules that use non-pointers attributes helps (and make sense) in some cases, as Dave and Mauricio illustrated.


It seems that for several data categories the only pointer attribute that is likely to be really use would be for the stand-off reference.


So, if we fill there is a need for clean up, maybe we should keep all global rules, but trim the pointers attributes in some cases.





From: Felix Sasaki [] 
Sent: Tuesday, October 23, 2012 9:46 PM
To: Yves Savourel
Cc: Dave Lewis;
Subject: Re: issue-51 too many global rules



2012/10/24 Yves Savourel <>

Hi Dave, Felix, all,

The rarity of use case for pointers in those data categories is also compounded when the data category has several values: As Felix noted, because of the complete overriding clause, we can only use all pointers or none for a data category.

I would still argue to keep any refPointer to stand-off markup though. XLIFF 2.0 is a use case for it.


+1. However, Mauricio had a nice example about global rules for provenance


That seemed elegant and is similar to Dave's previous , so I'm not sure anymore how to move forward about the issue. Thoughts?






-----Original Message-----
From: Dave Lewis []
Sent: Tuesday, October 23, 2012 5:37 PM
To: Felix Sasaki
Subject: Re: issue-51 too many global rules

Hi Felix,
I have a further thought actually on pointer and ref pointer attributes in general below:

On 23/10/2012 17:24, Felix Sasaki wrote:
> If nobody uses the expressiveness, we don't need to add it to new data
> categories in ITS 2.0. I still get nightmares from rubyPointer .... :)
> In ITS 1.0 the expressiveness was mostly used on a per format basis,
> e.g. saying "all 'alt' attributes at HTML 'img' should be translated.
> I don't see the "per document format" or even "per template" use case
> for
> QualityIssue, Quality Precis, Disambiguation, mtConfidence, text
> analysis annotation, translation provenance.
> So for these the "pointer attributes" (or even reference pointer only)
> might be sufficient.

So, I'm not even sure that we need even the pointer attributes for certain data categories.

I tried to outlined in:

that pointer and points ref attributes didn't make much sense for data categories that were more provenential in nature, i.e. the were generated in the localisation chain, rather than being internationalisation instructions from content authoring processes.

I probably didn't argue this very clearly, and apologies Felix for being slow in clarifying as you asked in:

Yves makes the point more succinctly in:

where he says:
"I'm less concerned with 'complex/rare' data categories like Disambiguation, or MT Confidence, because it's unlikely an existing format has the equivalent."

I'd agree. Certainly with provenance I found it difficult to come up with examples using Pointer and RefPointer data attributes. I couldn't think of an existing schema elements that I'd point to, so the examples use rather contrived elements. If this is the case, should we just state that people should use the direct value or ref ITS data attributes and drop Pointer and RefPointer in both GLOBAL and LOCAL usage?

Shaun's excellent point about what to do when more than one node matches the relative path of a pointer is also significant here in killing off

I don't know if this applies to the quality issue data category. They use Pointer for mapping to 'native' attributes in example80 in the current draft, but did the native 'issue' element and those attributes 'type', 'note', 'value' and 'profile' attribute reflect a known used schema?

So, Felix, regardless of the outcome of the other global rule discussions, in several cases (i.e. potentially for QualityIssue, Quality Precis, transAgentProvenance, disambiguation, text analysis annotation/confidence and mtconfidence) rather than pointers being 'sufficient', I don't think we need them _at all_ (for global or local).


> Best,
> Felix


Felix Sasaki

DFKI / W3C Fellow


Received on Wednesday, 24 October 2012 13:28:59 UTC