ID strings are used in a few places in the OTrunk code. These cause problems when adding new database types, and when copying objects. However using indrect references can be pretty annoying when writting otml directly.

It is possible to have the data layer automatically convert id strings. To do this it would need to know the pattern to look for these strings. and then when it finds them, it would need to add those references to standard map which a view of the object could later use to look up the actual object. If a portible version of the otml it should be processed and rewritten out.

From a scripting point of view it makes sense for references to objects in the document have local string tags. Then these can be used so the resulting components can be referenced by those tags. Otherwise the script has to lookup the components using object lookups which are tedious, hard to maintain, and limit the number of views of a single object to one.

The ideal solution would be to have the text of the document be a first class citizen of the OT system. Then elements in the document can be individually changed and those updates will be saved, as well as the abillity to embed objects directly in the text of the document. This is currently not possible with the otml we are using. But this should be possible with the move to emf.

So the plan is that we modify the existing otml parser to automatically convert references to objects in documents to use an indirect references. This will break the reporting tool. But it can be fixed. The otml written out will then have these modified references, and copying will work without special casing documents.

Then with the emf code we will attempt to add full support for xhtml inside of the emf document. Then in this case objects can be embedded directly inside of the document text. And references will be real references.