Hi Jeremy, > This seems overly complicated. [...] > There is an issue to do with loops in the definitions - I would have > thought that the simplest approach is to (weakly) prohibit them. i.e. > every definition is required to be grounded, where: Yes, this would be the obvious approach. In fact we thought about this, but we that we think it does not make sense, because... > a) if we know that URI u names graph g, then u is grounded > b) if we have a URI u that names a graph via query Q that depends on > graphs named u1, u2, ... uk and each of u1, u2, ... uk is grounded then > u is grounded. ...how should I prevent - in a distributed, uncontrollable environment like the Web - the owner of u2 from adding a view based on u to u2 tomorrow? Perhaps this does not even influence what I can infer from u, so I surely do not want the whole view to be false tomorrow just because of this change. Best regards, Simon
Received on Thursday, 18 October 2007 10:23:25 UTC