#305 Deprecation of data.key and data.word attributes

RED
closed-wont-fix
Martin Holmes
5
2012-11-28
2011-09-23
Martin Holmes
No

Over the past few years, there has been some discussion of what to do about attributes such as @key, which have "magic token" values. The general feeling has been that they should be deprecated in favour of attributes such as @ref which take data.pointer values, and that the Guidelines should be updated to show clear examples of how to use data.pointer when you don't have a reliable or practical URL to point at (essentially, when you actually do want a magic token). So this:

key="myThing"

could be this:

ref="local:myThing"

using a private scheme. Other attributes that ought to be addressed at the same time include:

@lemma (data.key)
@loc (1–∞ data.word)
@cRef (1–∞ data.word)
metSym/@value (1–∞ data.word)

all of which, I think, might be better served by @ref or @target.

There has been a lengthy discussion of this in Council recently starting here:

http://lists.village.virginia.edu/pipermail/tei-council/2011/013761.html

and a previous discussion minuted here:

http://www.tei-c.org/Activities/Council/Meetings/tcm45.xml

This FR is also related (although it would presumably be declined given the previous discussions):

https://sourceforge.net/tracker/?func=detail&aid=2919640&group_id=106328&atid=644065

On that ticket, Sebastian suggests:

a) actually use URNs in their full raging glory, as in
<p facs="urn:www.ox.ac.uk:P45">Hello world</p>

b) invent an arbitrary schema, which means something to you:

<p facs="oxpics:P45">Hello world</p>

c) use "tag URIs" as documented in RFC 4151
(http://www.ietf.org/rfc/rfc4151.txt)"

<p facs="tag:indiana.edu:P45:>Hello world</p>

This ticket invites preliminary discussion about:

- Whether we want to deprecate some or all of these attributes

- What precise recommendations we want to make to encourage people not to use them

- Exactly which attributes we should be concerned about

Discussion

  • Lou Burnard
    Lou Burnard
    2011-11-02

    • milestone: --> RED
     
  • Lou Burnard
    Lou Burnard
    2012-03-15

    I continue to believe there is/should be a place in the world for magic tokens. We just need to be clear about what we are doing when we use them in a file intended for interchange. The existence of a distinct attribute to specify them, which by definition has a value that is not necessarily a URI, seems very useful and appropriate. The question raised on FR 644065 has to do with the specific case of an attribute which is defined as *not* taking a magic token, but for which we would like to supply something like one, tricked out in some way as if it were a kosher URL. That surely is what URNs were invented for, so I see no problem in using them: in whatever syntactic form we can identify. I think it would definitely be a mistake to just "deprecate @key". It has its place. We just need to be clear and consistent about what that place is.

     
  • James Cummings
    James Cummings
    2012-06-29

    • assigned_to: nobody --> martindholmes
     
  • Martin Holmes
    Martin Holmes
    2012-06-29

    Note the recent proposal for a formal method of dereferencing magic tokens which are expressed through private URI schemes:

    http://wiki.tei-c.org/index.php/Private_URI_Schemes

    I think it would be a good idea to propose, simultaneously, that:

    - Private URI schemes be used in place of traditional magic tokens, and
    - they be dereferenced using the method described in the link above, so that they are no longer "magic".

     
  • Piotr Banski
    Piotr Banski
    2012-09-21

    @lemma is a good candidate for the status of a magic token. It isn't always that you have the luxury of pointing to a dictionary that describes the given lemma. Sometimes, you just want to cite it straight in your annotation, while being aware that in <w lemma="go">went</lemma>, "go" merely looks like a natural language string -- it is actually an identifier of, let's say, a dictionary entry. Except that in a linguistic corpus, you get thousands of lemmas. Expecting there to be a multi-thousand-entry dictionary into which you can conveniently point from each <w> is not (always) realistic.

     
  • Piotr Banski
    Piotr Banski
    2012-09-21

    Now, if you tell me that all it takes is changing the previous example into e.g. <w lemma="dict:go">went</w>, my answer would be that I myself see no problem with this, but I'm concerned about the existing extensive linguistic resources that use @lemma in a magical way.

     
  • Kevin Hawkins
    Kevin Hawkins
    2012-09-21

    ... in which we settled on using "tag URIs" instead of private URI schemes.

     
  • Martin Holmes
    Martin Holmes
    2012-11-28

    • status: open --> closed-wont-fix
     
  • Martin Holmes
    Martin Holmes
    2012-11-28

    Per decision of Council in Oxford, Sept 2012, this ticket is to be closed, and is replaced with a newer, shinier ticket at http://purl.org/tei/bugs/3590818.