On Wed, Jun 20, 2001 at 07:13:26PM -0700, Thomas Bushnell, BSG wrote: > Hubert Chan <[EMAIL PROTECTED]> writes: > > > But for the situation we are talking about, they would need to have the > > same interface, since a PGP front end needs to interact with the PGP > > program. So in the PGP front end depends on the "pgp implementation" > > virtual package, but the PGP program doesn't have an interface that > > works with that front end, you get a broken distribution. > > No, you're wrong. The mailcrypt front end, for example, works with > both. And that's the case we are talking about.
It depends how the compatibility works. If it's mailcrypt providing the compatibility, then it's mailcrypt that should list the compatible packages, IMHO. Depends: implies more than just something-like-that-must-be-present, it implies a particular supporting interface is required for the package to work. If OTOH mailcrypt is just using a small subset of the commands such that any reasonable PGP clone should work with it, then the virtual package is possibly the right way to go. It sounds to me like a bad cure to the wrong problem, policy should just be clearer that a main package is allowed to depend on an |ed set of packages providing at least one is in main, IMHO. -- Colin Phipps PGP 0x689E463E http://www.netcraft.com/
