On Thu, 11 Dec 1997, Weibel,Stu wrote:
> I hope I am not being premature (though, I think we should be aggressive
> about putting this issue to bed), but it seems to me that the
> combination of legacy experience, current implementation, and common
> concensus suggests that the best recommendation for delimitting multiple
> fields in a tag is the semicolon. (As John Kunze pointed out to me, this
> was also the conclusion that that the original User-Guide group
> proposed).
>
> This does not pre-empt other scheme-defined approaches that might arise,
> but rather proposes a default that is already in wide use, and further
> suggests a practical complexity threshold: if doing this results in
> apparent ambiguity, then the implementors are well advised to either use
> seperate fields or to adopt a formal scheme.
>
> Any volunteer to write up the one-page recommendation for the DC web
> site?
>
> stu
I'd be happy to volunteer for this one, if we _do_ have consensus on the
following 2 issues. I'm assuming we're all OK with ';' as the delimiter.
1. Within which elements, and sub-elements, does ';' play
a delimiting role?
a) All elements (and all sub-elements?)?. [By consensus unworkable]
b) Just DC.subject and all it's subelements (if it has any -I forget)
c) Just DC.subject and none of it's subelements.
d) None of the above: a selection of elements to be agreed[when?]
Comments from the sub-element working party from DC-5 should be able
to settle (b) -vs- (c). I would expect ';' to be safe in all
conceivable SCHEME-free sub-elements of DC.subject, ie. (b) is the best
option here.
2. The effect of SCHEME
a)
As a general rule, ';' only signifies a delimited field in the
absence of a named SCHEME. It might well be that various SCHEMEs
attach similar meaning to ';' or other characters. To avoid
conflicts derrived from other uses of ';', Dublin Core processing
software should not assume without good reason that ';' has any
special significance in elements that have a SCHEME.
b)
';' always splits fields, SCHEME or no-SCHEME
c)
Neither of the above. Whilst recognising that ';' splitting could
damage some SCHEMEd data, we could still reasonably want a
semicolon delimited list of -say- Dewey Decimal classifications. For
long lists of short numbers, this is very desirable.
eg:
<META NAME="DC.subject" VALUE="123.321;32.123;100.101" SCHEME="DDC">
Suggestion:
Go with (a) but recognise (c) as a requirement when/if we get round to
specifying machine readable descriptions of fields that draw on standard
classification schemes. (a) does allow the _possibility_ of ';' delimited
SCHEMEd fields. But it does not suggest any mechanism for telling us
_when_ we can be sure that a SCHEMEd field deserves to be chopped into ';'
delimited bits. (This may be another one for the RDF wishlist...)
Am I right in thinking that 1(b) and 2(a) describe something approaching a
consensus?
If not I'm sure I'll find out soon :-)
Dan
--
[log in to unmask]
Research and Development Unit tel: +44(0)117 9288478
Institute for Learning and Research Technology http://www.ilrt.bris.ac.uk/
University of Bristol, Bristol BS8 1TN, UK. fax: +44(0)117 9288473
|