Fri, 27 Jan 95 12:56:48 -0500
Not exactly. You must invoke the outermost keeper that could have been
responsible for the fault. In your example, since segment A's keeper
doesn't intend to play tricks with segment D, segment A has a read/write
key to segment D. Therefore A isn't responsible for the fault. D is
because D contains a read-only key.
That sounds like it works for write access faults. Is there an
analogous trick that can be used for handling invalid data, or does
that really have to go all the way through the chain?