RE: AW: [all] revision to disambiguation text

Hi Tadej,

This helps a lot.
So we really have:

* Either (target type class):
   * disambigClassXYZ mandatory
   * disambigGranularity optional
* Or (target identity):
   * Exactly one of:
      * disambigIdentXYZ + disambigSource
      * disambigIdentRefXYZ
   * disambigGranularity optional

I'll update the specification so we don't have the possibility to define only a selector.

Thanks,
-yves




-----Original Message-----
From: Tadej Štajner [mailto:tadej.stajner@ijs.si] 
Sent: Tuesday, October 23, 2012 5:31 AM
To: public-multilingualweb-lt@w3.org
Subject: Re: AW: [all] revision to disambiguation text

Hi, Yves,

On 22. 10. 2012 23:59, Yves Savourel wrote:
> Hi Tadej, all,
>
> I'm afraid I'm still a bit lost in the Disambiguation data category.
>
> I now got the two types of addressing (source+Ident, or identRef) as described here:
>
> "When using a disambiguation rule, the user MUST use one of the use cases for disambiguation: specifying the target type, or specifying the target identity. For the latter, the user MUST use only one of the two addressing modes:
>
> - Using disambigSource and one of disambigIdent or disambigIdentPointer to specify the collection and the identifier itself.
> - Using one of disambigIdentRef or disambigIdentRefPointer using a URI for the disambiguation target."
>
> But this either/or is for the second type only (target identity case).
>
> So:
>
> - what attributes are used for the first type?

With the recent revisions, the phrase 'specifying the target type' would sound clearer if it said 'specifying the disambiguation target type class'. In this first use case of the disambiguation rule where we just want to specify the type, the user would use:
* disambigClassXYZ mandatory
* disambigGranularity (optional - used if it cannot be inferred from the value of disambigClassXYZ)


- how the first and second types are actually represented in the first place? Currently both disambigClassXYZ and disambigGranularityXYZ are optional, so technically we could have a rule with just one of the information for the addressing.


When specifying the type class (as before)
* disambigClassXYZ mandatory
* disambigGranularity optional

When specifying the identity:
* exactly one of:
   ** disambigIdentXYZ + disambigSource
   ** disambigIdentRefXYZ
* disambigGranularity optional

>
> Looking at this: "...specifying the target type, or specifying the target identity" I'm guessing the following:
>
> The 'target identify' case is when you use disambiguClassXYZ with 
> disambigGranularity='entity' (?) The 'target type' case is when you 
> use disambigGranularity with one of the concept (?)

The 'target identity' case is where one of the disambigIdent(Ref)XYZ attributes is used, regardless of granularity value.
The 'target type' case is where one of disambigClassXYZ is used, regardless of granularity value.

The granularity applies to both identity and class. It's not what I would call orthogonal with respect to class and ident values, but it is orthogonal with regard to addressing mode and use case.

I hope this clears it up? I can come up with a combination matrix if necessary.
-- Tadej

>
> Or is it something different?
>
> Thanks,
> -yves
>
>
>

Received on Tuesday, 23 October 2012 16:58:13 UTC