- From: Graham Klyne <GK-lists@ninebynine.org>
- Date: Mon, 01 Mar 2010 12:30:34 +0000
- To: Jakub Kotowski <jakubkotowski@gmx.net>
- CC: public-cwm-talk@w3.org
Jakub Kotowski wrote: > Graham Klyne wrote: >> FWIW, I ran into some similar problems when playing with Swish and >> inference rules - handling bnodes got rather messy. > Why were bnodes a problem? Because of scoping issues - mainly (IIRC) in backward-chaining rules that introduced new existentials. It turned out that I needed these newly introduced nodes to be scoped to the particular new goal clause(s) generated by the rule, but bnode scoping is, by definition, scoped to the entire containing graph. (I stalled about that point thinking about mechanisms to create an implementation for subgraph scoping of bnodes, but it all seemed over-complex. Later, the aforementioned idea of converting all existentials to unique identifiers as they are encountered occurred to me, but by that time I'd moved on.) #g
Received on Monday, 1 March 2010 19:05:45 UTC