On Sat 21-Aug-2004 at 14:34 -0500, Eric Wilhelm wrote: [snip hub and node model] > In this scenario, how are the rhizopod-format databases affected by a > format-incompatible upgrade to the 3D relational modeling program? They > aren't. That's correct so long as applications are always importing and exporting to and from the format rather than accessing the data directly. The many-files-in-a-directory thing is a distributed system, the two major characteristics are: 1. Decentralisation, data can be manipulated at a low level without tools needing to touch the entire dataset or locking it. The guiding design principle should be that, if information can be delegated to the lowest level, then that's what should happen. This is why I think that each element should have it's own units, version, copyright etc.. fields - Even if certain applications will always set this data identically for every element - Even if ultimately these fields just reference other locations. 2. Addressability, every item of data has a persistent location. "Drawings" can be referenced by pointing to a directory (useful for xref/block/symbol/viewports). "Elements" can be referenced by pointing to files (eg. the address of a layer definition is the location of a file containing that definition). "Attributes" of elements need to be referenced using a new syntax. I'm suggesting repurposing the URI Fragment Identifier syntax, so the coordinates of the first point of a line can be found at "./path/to/file-containing-line.yml#points/0". Sorry if I keep banging on about these things, I hope I'm adding something useful to the mix. -- Bruno