In message <[log in to unmask]>, Martin Holmes <[log in to unmask]>
>I think that's rather an extreme position.
Thank you ;-) I'm not being deliberately irritating, honest: I'm just
trying to tease out an issue which I think is there.
I agree that the existing mechanisms "do the job" in many cases, and I'm
not seriously suggesting that they be scrapped. However, I still wonder
whether links to external entities/resources deserve a little more
support than they currently get.
When I make a link out to an external resource, I might want to state:
- what type of resource it is I am linking to;
- where that resource is;
- what relationship exists between the text I have marked out and the
resource I am linking it to (i.e. _why_ I am linking to it)
in addition to the resource's actual identifier.
The first of these can be dealt with by the type attribute, subject to
it not being hijacked to represent the type of reference, as previously
The second, you might argue, should be part of the identifier, i.e. the
value of the key attribute. In a URI context this would undoubtedly be
true, since all you need is an absolute URI, possibly with a part
qualifier. However, in a database context you might validly want to
record the database and/or table and/or field in the context of which
this identifier is valid. I don't necessarily want to make this part of
the key attribute itself, nor do I want to be forced to infer it from
the type of resource. (You might have two distinct person authority
The third I don't see any place for within the current scheme of things.
Yet if we are wanting to generate e.g. RDF triples or Topic Map
associations from our TEI links, this type of "predicate" information
would be very useful.
>The system as it stands has a lot of flexibility; there are
>shorthand/syntactic sugar elements such as <placeName> that are useful
>to lots of people, and which also help newcomers get started without
>too much difficulty, and there are more general options such as
>name/@type and [log in to unmask] You could even use <seg
>corresp="#joe_bloggs">, which could point at something which defines
>its type with great precision. <ref> can also be used to point to an
>internal or external real-world entities, assuming URIs can find them.
>Feature structures, while extremely powerful in some contexts -- I've
>used them in a linguistic dictionary project, where they were really
>essential -- are overkill for many projects.
>I agree with Lou that this is really an issue for project
>documentation. Which of the many alternative methods any project
>chooses for management of names and referring strings will depend on
>the nature and goals of the project, and as long as the tags aren't
>abused and the documentation is clear, I think the existing system
>Richard Light wrote:
>> In message <[log in to unmask]>, Lou Burnard
>><[log in to unmask]> writes
>>> It's true that TEI currently has nothing for <ship>s and requires
>>>you to use <org> (<orgName>, <rs type="org">) for families,
>>>institutions, peoples, or collectives. Though obviously generic
>>>elements like <name> and <rs> can be used for them.
>> The discussion on this topic possibly points up the lack of a
>>coherent ontology of things-referred-to. Ships, for example, can be
>>seen (at least in one sense) as a sub-class of objects. The CIDOC CRM
>>springs to mind as a potential organising principle. This is
>>possibly an area that the Ontologies SIG could look into?
>> Having started by asserting that <placeName> etc. should be allowed
>>to contain anonymous names, I'm now starting to think the opposite:
>>that we should abandon all the "naming" elements (including <rs>) to
>>their original purpose and think about having an element whose sole
>>purpose is to link out to external real-world entities of relevance
>>to the text.
>> I would argue that even <rs> is compromised for this purpose, since
>>it is constrained to contain "a general purpose name or referring
>>string". What if a whole paragraph is about a specific person? What
>>if there is no particular phrase on which you can hang your external
>>link, and you want to put it in as a "milestone"? Lou has already
>>commented on the ambiguity of what the <rs> "type" attribute is meant
>>to be typing.
>> Should we be simply switching to the use of feature structures for
>>this type of external reference? (That's what we have done at the
>>Wordsworth Trust, as we will be mentioning in our talk on Friday.)
>No virus found in this incoming message.
>Checked by AVG - http://www.avg.com
>Version: 8.0.175 / Virus Database: 270.8.6/1768 - Release Date:
>11/4/2008 9:38 PM