[
Lists Home |
Date Index |
Thread Index
]
james anderson <james.anderson@setf.de> wrote:
| Arjun Ray wrote:
|> A name mapping is always necessary, between the arbitrary names an
|> author could use and the "well-known" (or externally predefined) ones
|> he may wind up using.
|
| it appears that we do not agree, that the mechanisms which you describe
| would in general be sufficient to address documents of the form of the
| last example.
The last example is no more and no less arbitrary than the others. That's
the point: you start from an arbitrary configuration and proceed to map
vocabularies. It isn't necessary that an author actually visualize this
as what he's doing; it's only what he actually winds up doing anyway.
If you're asking whether the control information needs to be physically
present in the undiscriminated form of the document, the general answer is
no. You could do it through an ancillary process such as an LPD, or an
instance specific XSLT "transform" or whatever. (Using an internal subset
one could use the much maligned attribute defaulting method to save on
markup in the instance. There are many ways to skin this cat; the fact
remains, though, that *some* mapping - even a trivial identity one - is
inherently necessary in associating an external vocabulary with the native
element structure of a document.)
| namespaces are specified as a means to ensure that graph labels remain
| unique under arbitrary graph combinations. the techiques you describe
| appear to be intended to effect arbitrary graph relabeling.
Actually, (sub)graph labelling, not graph relabelling. The original graph
has no scrutable labels, only structure.
| neither is sufficent for all cases.
Could be. The discussion isn't over. I wish it had *started* years ago.
|
