[sc34wg3] Towards TMDM 3.0

[Rani Pinchuk]

Dear Lars,

Thanks for taking your time with this.

Let me just summarize:

Item identifiers are used only in order to identify items. Item 
identifiers of topics representing different subjects should not be 
equal. This is true universally (since the item identifiers include the 
IRI). However, item identifiers of topics representing the same subject 
may be the same.

Usually we will not merge by item identifiers at the first place, as 
usually they will be different (even if they are proxy to the same 
subject) - just because the IRIs of different topic maps are different.

Let's assume that we merge topic A and topic B because of any reason 
(equal PSI, manually merge...). We collect the two item identifiers of A 
and B. The fact that we collected the item identifiers might help us to 
merge to a third topic C.

How exactly? This topic C will not have the same item identifier as A or 
B (as the IRI is different). How could we possibly use this collection 
of item identifiers for future merges?

Thanks again,

Rani

Lars Marius Garshol wrote:
> * Rani Pinchuk
>> I am sorry to persist, this is still not clear to me.
> 
> If you don't understand I think you *should* persist. :)
> 
>> If I understand correctly, you say that usually item identifiers are  
>> not
>> used as subject identifiers, but when there are no subject identifiers
>> we have to have something in order to merge, so we use item  
>> identifiers.
> 
> Almost. I'm saying that item identifiers are usually what we use to  
> identify topics, because usually there are no subject identifiers. The  
> rest is right.
> 
>> I find this dangerous. Many topics might be merged by mistake (some  
>> IDs
>> might be popular defaults given by the Topic Maps engine. "1" for
>> example. Other might be simply confusing choices. "bank" for example).
> 
> See below.
> 
>> I would actually prefer to merge by names which seems to me much more
>> safe. Or, to use more sophisticated algorithms for merging (e.g. using
>> NLP etc.).
> 
> Merging by name just isn't safe. Of course, you can do it sometimes,  
> and sometimes you can use NLP, but we can't require this sort of thing  
> in the standard.
> 
>> I understand from what you write that ID in XTM is an item identifier
>> without the topic map IRI. Is this correct?
> 
> Yes.
> 
>> At first, I thought that the item identifier includes the topic map  
>> IRI,
>> and therefore you could claim that the item identifiers are unique.
> 
> Item identifiers are absolute IRIs, so that if you download from
> 
>    http://www.garshol.priv.no/tmp/tst.ctm
> 
> the file
> 
>    lmg - "Lars Marius Garshol".
> 
> the item identifier for the topic will be
> 
>    http://www.garshol.priv.no/tmp/tst.ctm#lmg
> 
> (If you think this conflicts with my previous "Yes" above then please  
> read your question one more time. The *ID* is not an IRI. The item  
> identifier is.)
> 
>> All in all, I find that the model you describe above is very  
>> confusing:
>> item identifier actually identifies subject, but on a lower grade then
>> subject locators and identifiers.
> 
> The TMDM says: "An item identifier is a locator assigned to an  
> information item in order to allow it to be referred to." (Clause 5.1)  
> So it really does identify the item, but ... well ... indirectly that  
> does also identify a subject and we can't do anything to prevent that.  
> But it really does refer to the item, and not what the item represents.
> 
> If you look at the item identifier above I think you'll agree that
> 
>   (1) Taken literally, it refers to the "lmg" topic.
> 
>   (2) It wouldn't be a huge stretch to say that in some sense it also
>       indirectly refers to me.
> 
> --Lars M.
> http://www.garshol.priv.no/blog/
> http://www.garshol.priv.no/tmphoto/
> 
> 
> 
> _______________________________________________
> sc34wg3 mailing list
> sc34wg3 at isotopicmaps.org
> http://www.isotopicmaps.org/mailman/listinfo/sc34wg3

-- 
Rani Pinchuk
Project Manager
Space Applications Services
Leuvensesteenweg, 325
B-1932 Zaventem
Belgium

Tel.: + 32 2 721 54 84
Fax.: + 32 2 721 54 44

http://www.spaceapplications.com