Re: [Proposal] DTM separation work

[Joseph Kesselman]

> None of the DTM interfaces are touched by this change (of factoring DTM out)

Granted... but *I* may want to change the DTM interfaces significantly, and the result may not be something we really want other folks using. I'm proposing we abandon node handles. That means abandoning most of the current official public APIs of the DTM, and using the node-index based calls instead... and leaving most of those as non-public, with the cursors being the new public API. It isn't clear that the resulting revised DTM -- a specific back-end data model with no supported public APIs -- merits having its own jarfile.

(Especially given how much trouble we already have explaining classpath mistakes...)


I may be being excessively paranoid. But you did ask for opinions, and mine is that
this is not the moment I would choose for a dtm.jar effort, since the outlines of that module may be in flux and thus the benefit of a separately reusable DTM is unclear.

On the other hand, cleaning up the dependencies between levels of Xalan hierarchy strikes me a great idea.



______________________________________
Joe Kesselman  / IBM Research