[tei-council] a proposal for a change to ODD (copy of ticket I just put in SF)

James Cummings James.Cummings at oucs.ox.ac.uk
Tue Mar 9 05:28:30 EST 2010

Sebastian Rahtz wrote:
> re: repository of examples, I don't see that we need do anything
> to let people use it. we have <xi:include> which can do the job fine,
> no?

Yes, or just a ptr would be fine.  The question is more about 
where it is pointing.  As you are aware I think that the examples 
should be decoupled from the Guidelines prose and the element 
specs and stored separately.  This way two places in the 
Guidelines could transclude/xinclude the same example, and when 
the example was updated, both places would be updated.  One of 
the fundamental principles of easy re-use, no?

>> I'd be tempted to say that an ODD should maybe check simply that all the 
>> elements in an example are present for the current schema, and if not 
>> add a comment to the example indicating it might be out of date because 
>> element <foo> is no longer included in the schema.  That seems like it 
>> would be much more straightforward to program, doesn't remove the 
>> example, but warns the user that it may be inaccurate.
> that doesnt make me _very_ happy. "we've found a problem but we cant
> fix it, so hey just have it anyway" seems unfriendly. but I agree, programmable.

I was thinking of it more like "Warning: The following default 
example may not match your schema." but I see your point. I think 
having the warning is _better_ than providing the invalid example 
without a warning.  Not as good, of course, as providing a valid 
example, but I don't see how to do that without a great deal of 
complexity. If we're only testing element presence or not, rather 
than content models, then it is a rough estimation at best. (I.e. 
we know if it is completely invalid, it contains elements which 
don't exist in the schema, but we don't know if that attribute is 
allowed here or if the content model for this element has been 
changed.) The choices, as I see them are:

a) Provide the invalid example, no warning
b) Provide the invalid example, with warning
c) Do not provide invalid examples, reference just has no example
e) Attempt somehow attempt to find a valid example which includes 
this element in the corpus of examples, falling back to one of 
the above otherwise
f) Attempt somehow to validate the examples and use their full 
validity to do one of the above.
g) Do something else that is even better.

Of all of these b) actually seems the friendliest to me.


Dr James Cummings, Research Technologies Service
OUCS, University of Oxford

More information about the tei-council mailing list