Allow specifying packages are successors / incompatible

[jappeace]

Describe the feature request
Recently I forked the memory package as ram. This to end the basement/baseplate discussion by just ripping out basement from memory and other packages.

Now downstream users may get really bad error messages if they upgrade (crypton) for example. A type used from memory is incompatible with a type used from ram.
What I would like is to inform users that if the solver pulls in both memory and ram, something is going wrong. Instead of random compile or constraint errors.

This should make the community a bit more resilient against contactable but hostile maintainers.

Additional context
I decided to split it in this discussion: https://discourse.haskell.org/t/fork-basement-as-baseplate/12415/64

[geekosaur]

You might want to open a related issue against hackage-server, on the grounds that it'd be even better if Hackage reported these incompatibilities as well.

[phadej]

What I would like is to inform users that if the solver pulls in both memory and ram,

There is no general technical reason to think that depending on memory and ram (or microlens and lens, or cryptohash-sha1 and crypton) or any other pair of two package is "bad" configuration. Maybe suboptimal, but IMO none of package authors should be able to say anything about other packages.

Also in this particular case, memory may become maintained in some time in the future. Say, if Hackage policy on package (or rather package name) takeovers changes.

Don't try to solve social problem with a technical solution 👎

[phadej]

If anything, you could lobby that e.g. Stackage should simply remove memory from its snapshot, and forbid its inclusion. That will be a "social" solution.

AFAIK, so far there haven't been precedent for Stackage for rejecting inclusion of any package for other than technical reasons (i.e. doesn't build, doesn't update deps etc.(

[jappeace]

Cabal bans installing multiple versions of the same library. This is a similar situation.
But rather asking for an outright ban I'd be happy with a helpfull message indicating something is going woefully wrong (somehow).

Also in this particular case, memory may become maintained in some time in the future. Say, if Hackage policy on package (or rather package name) takeovers changes.

Then these packages should still not be compatible.