Node size doesn't matter

If you look at what Michael has done with GODDAG and consider his background (with SGML, TEI, etc.) it makes perfect sense that his approach would be with things like CONCUR and Extended XPath [EXPath].

Combining ideas from GODDAG, or more properly, ideas from EXPath, provides a way of referring to points within a text, spans, and overlapping spans, by way of graphs with a modicum of document structure polyancestry. By coincidence, the paper I'm currently writing (which may never see the light of day) has an introductory section comparing the paradigms of linking vs. mapping. What Michael seems to be basically doing is mapping, and while I've not looked into what Graham had proposed, I'm guessing it was mapping a document.

Since Topic Maps are designed precisely for this purpose, and have their own inherent graph structure (as opposed to a tree structure), both a Topic Map document and the territory-document it is mapping can have a mirrored graph structure.

Now, NODAL might seem somewhat orthogonal to this, but only a bit. NODAL is similar to (and influenced by) the Reiser File System by Hans Reiser (now commonly used on linux machines), which rather than create a file system upon a tree structure, uses node-level metadata to create a journaled graph system. Very high performance, I've been using it for three or four years now.

If one considers that an implementation of NODAL (as described to me by Lee) effectively breaks down the barriers between documents within a file system into what in the HyperText world are typically called "lexia" plus added metadata, if such a system were implemented across machine boundaries one would effectively have an enormous grove, i.e., with the right tools the Web would become one big grove. I think that's what Lee's thinking, anyway. He might be interested in this conversation. And if one considers what ReiserFS and NODAL are doing, it's the micro level to what a digital library is doing at the macro level. If we ignore document boundaries, there's almost no difference conceptually.

Eliot Kimber (coauthor with Steve Newcomb of the ISO 10744 HyTime standard, which provides us with the concept of groves) has a simplified grove model for XML called XIndirect which, coupled with NODAL, provides almost everything:

• XIndirect: Indirect addressing for XML
  W. Eliot Kimber, Extreme Markup 2003.


• XML Indirection Facility,
  W.Eliot Kimber, W3C Note 12 June 2003.