On 16/01/2012 06:12, David Moner wrote: > A possible problem I can envision is that this opens the door to the > creation of invalid archetypes without the possibility of validating > them at design time. > > A quick and dirty example just to get the idea. In an archetype, the > HISTORY.origin is fixed to "T10:00:00", a child EVENT.time is fixed to > "T18:00;00", and finally the EVENT.offset is fixed to "PT5H". Offset > is clearly not valid regarding time and origin, but we do not know it > since the expression to calculate it is not explicit anywhere (only at > the RM specifications). >
that problem can occur even with just two normal attributes, if they happen to have a mathematical relationship... - thomas