Re:Design questions for future filter model.

Alexandre Rafalovitch writes:
 > At 1:13 AM 20/8/96, Anselm Baird-Smith wrote:
 > >Alexandre Rafalovitch writes:
 > <snip>
 > > > It would be great if filters could register their requirements on filters
 > > > they are executed before and after.
 > > > Using any other not so flexible system (eg. Alphabetical) would bring
 > > > problems. (as people with Machintoshes know too well :-{ )
 > <snip>
 > >It should be easy, with the new filter design to have a resource
 > >maintain its *own* set of contraints on filter execution, since in the
 > >new design, it is up to the resource to ultimatley decide in what
 > >orders  the filters are to be called at perform time (again, for
 > >lookup, you will not be able to affect the above order).
 > 
 > I cannot see how this solve a situation I had in mind. What I am refering
 > to is: I create a new filter which I know should be executed before filter
 > A, but after filter B. The filter is written much later then any resource
 > serving my files. So, how would I make resources know in which order to
 > execute my filter? Would I have to go and edit each resource? Would they
 > have some generic algorithms built in? I just don't see how it would
 > happen.
 > 
 > Maybe I don't understand something and I can wait until release to see how
 > it works, but I felt that a bit more flexibility in filter execution order
 > would be only for the advantage without bringing any harm(?).

I am not sure I understand your point enough to do something at this
point. However, I know for sure that more flexible (or powerfull)
filters ordering can be added on top of the to-be-released API,
without trouble (eg by adding some sort of priority attribute to
filters, but I am really scared by going along this way)


Anselm.

Received on Thursday, 22 August 1996 21:14:44 UTC