In our Book Trade Correspondence project, we have a preference for using
W3C schemas. The background of the question of my colleague did not so
much concern the notation of the reference to the schema. The question
is more if it is possible to refer to a canonical schema (a schema
maintained at the tei-c.org website) in the first place.
We have already generated a schema ourselves using Roma, and we made
sure that this schema contains all the modules that we need (namesdates,
textstructure,transcr, amongst others). This is a schema which we
currently manage locally
(http://www.let.leidenuniv.nl/wgbw/DMT/tei_BTCP.xsd). But in our
opinion, it would be preferable if we can refer to a schema which is
accessible from the TEI-website, similar to way in which we used to
refer to the P5 DTD at tei-c.org.
I have found a number of W3C schemas at
when I try to validate one of our TEI documents using the
"tei_allPlus.xsd" schema, for instance, this produces various kinds of
errors in TEI, which appear to concern the schema rather than our TEI
file. The schema is invoked as follows:
The first error message that Oxygen produces is "src-resolve: Cannot
resolve the name 'xml:lang1' to a(n) 'attribute group' component. Is
there perhaps a problem with the way we invoke our canonical schema?
Any help with this would be greatly appreciated.
From: TEI (Text Encoding Initiative) public discussion list
[mailto:[log in to unmask]] On Behalf Of James Cummings
Sent: woensdag 25 augustus 2010 23:48
To: [log in to unmask]
Subject: Re: DTD > Schema
On 25/08/10 18:52, Sebastian Rahtz wrote:
> Sort of. If you use RELAX NG, then no. The closest is the notation
> <?oxygen RNGSchema="testmav.rnc" type="compact"?> but thats just a
> convention they made up. The RELAXNG massive don't believe in the idea
> of a 1:1 correspondence between XML file and schema, as they think you
> want to validate against different schemas at different times.
For what its worth I think this is the right decision and a good thing.
XML Processing Instructions are very useful for this kind of thing.
There is nothing saying you couldn't have 5 processing instructions all
for different types of editor or processor and none of them would
interfere with each other. I don't find it a bad thing for these to be
different (or indeed point to different schemas). Obviously, I also
agree with that Relax NG crowd that you want to validate against
different schemas at different points in your workflow and/or project
> Frankly, I think
> the notation was a dumb idea from the W3C.
It certainly doesn't seem like the best idea.
James Cummings, Research Technologies Service University of Oxford