On Sat, Jul 25, 2009 at 7:23 PM, Barry
Leiba<barryleiba(_dot_)mailing(_dot_)lists(_at_)gmail(_dot_)com> wrote:
Use relative URIs to represent resources managed transparently by the
Sieve Engine.
This was my thought as well, when Alexey first brought the idea up.
Originally, the idea was that the list name was just a string,
understood by the Sieve processor. It quickly became clear that that
was problematic, and we should go with URIs. Some thought they should
always be absolute, and I thought relative ones should be allowed. We
batted that around for some while, and settled on absolute.
Apart from the reason you mention -- that it can be confusing -- the
main argument against using relative URIs is the difficulty in clearly
(and interoperably) defining what the relative ones mean. Saying that
the Sieve processor understands what they mean ignores the real
problem with making sure the user (and/or the client) also understands
it.
I think we should stay with absolute URIs.
the location and interpretation of resources with arbitrary semantics
across arbitrary protocols is - i think - too difficult a problem to
be tackle by this draft. therefore, we should be satisfied now just by
specifying what we can and leaving these difficult interoperability
problems to other expert groups.
- robert