Cart
[Home] [By Thread] [By Date] [Recent Entries]
> > > 2. anyone can give a rundown on why they went for a diff technique > > rather than a direct pat specification with a replacement value? No but here is a different approach that may be of interest. For a commercial client I implemented a system that just used a dirty flag on elements (and attribute giving clean/dirty/new.) Their requirement was in a similar area to updategrams: they needed to extract trees from a database, update sections and then they just wanted the updated elements only merged into the database. Because we used column=attribute conventions for the database XMLification, the dirty bit was in fact dirty rows not dirty fields: for their application the coarse-grain approach was fine. It seemed out that pruning the XML data of empty data values (or data=(number)0, or null) was easier for performance (they had lots of unused columns) from the user perspective than fine-grained update, but I don't expect that is a general rule. This is not nearly as sophisticated as the updategrams, of course, but it works well as far as it goes. Rick Jelliffe
|
