Actually here is a description of what we are using. We are currently
using EPN and XPointer within two different applications.
Multilingual alignments :
A first release of our tool for aligning source texts with their
translations was based on the EPN and we simply used the ID keyword. We
also used the link, xptr and linkGrp elements to group alignments from a
We are rewriting this tool to become fully compatible with the XPointer
and XLink WDs. The main pb is coming from the use of the doc attribute
(of type ENTITY) within a linkGrp element. There is no mechanism for
factorizing a target URL to avoid its repetition. Something like this
would be nice :
<xptr id="x1" xlink:from="id(d1p1s1)">
<xptr id="x2" xlink:from="id(d1p1s2)">
<link targets="#id(d1p1s1) x1">
As u can see, we have modified (or extended) the type of the 'targets'
attribute. It contains a list of Locators and not simply a list of
IDREF. We could also avoid the use of the xptr elements :
<link targets="#id(d1p1s1) id(d1p1s1)">
The second XPointer (without the '#') is a reference to an element
within the 'base' document. There are certainly some other solutions...
TAG annotations :
In this application (based on our TagML specification, out asap), we are
doing an intensive use of the XPointer specification. The keywords we
are using are: id, child, descendant, ancestor, psibling nsibling.
[log in to unmask] said:
] - How do people feel about changing the attribute names (currently
] 'doc', 'from', and 'to') to match those defined by Xlink?
1/ I am not sure that it is useful... The way we could use 'doc', 'from'
and 'to' attributes is rather different that the way to use XLink
atributes. I think that both attribute sets can cohabit together.
2/ XLink attributes do not need to be declared. They are/will be
considered as 'global XML attributes'.
[log in to unmask] said:
] (Personally, I *hate* the name 'href', but I'm willing to be in the
] minority, and I might have to learn to live with it in any case.)
Same as me.
I am also rather disappointed with the XPointer STRING keyword.
I think that there should be a 'normalisation' between the EPN syntax
and XPointer's. If some projects are currently using the less widespread
keywords (HYQ, FOREIGN...), there should be a special TEI extension to
the XPointer specification to take into account these keywords.
Personally, I am considering the use of the TOKEN keyword and even STR
from the TEI extended pointers. Could it be possible to keep them please?
I guess I was clear enough...
[log in to unmask] | Office : B.228
http://www.loria.fr/~bonhomme | Phone : 03 83 59 30 52
* Serveur Silfide : http://www.loria.fr/projets/Silfide