Revise deprecation clause (close #12)

[sol]

(as suggested by @hvr)

[chreekat]

Anything I can do to help move this along?

[hvr]

@chreekat Well, I did already float this to the CLC over a year ago at https://groups.google.com/forum/#!topic/haskell-core-libraries/HCAepXp0NKg but there wasn't a clear resolution yet as the CLC members didn't speak out. So he next step would be to try to get the CLC members to express an opinion, by writing a mail to the haskell-core-libaries list. If there's no objections from the CLC, we can move on to announce this to the community wide libraries list and maybe CC to haskell-cafe, and after say 2-3 weeks implement it and release it as part of "PVP 1.1"

/cc @ekmett

[sol]

That sounds like a lot of pain to get changes into the PVP. That would not be a problem per se if the PVP would already represent some sort of community consensus. But that is not the case. At least parts of the PVP are highly controversial.

[chreekat]

@hvr I see that Kmett is on the CLC and gave a thumbs up in #12 itself. Plus, the conversation in the thread you linked looks entirely positive. You yourself are a Hackage trustee. Can you help clarify what a PR on this repo needs in order to be merged?

Unlike @sol, [this was referring to "This would not be a problem per se", but that's not enough to really judge someone's perspective] I think nailing down the process and making it efficient is actually pretty important for open source, volunteer projects. I have a little bit of experience in that sort of thing, which is why I am trying to be of assistance. If I need to send another email to the CLC I want to ask the right questions. However, I suspect it isn't actually necessary. (I also have no opinion on the merits of the PVP itself.)

[hvr]

@chreekat well, let me try to reach out to some CLC members directly; I'll report back (and thanks for wanting to help out!)

[chreekat]

Ok! Sounds good.

[RyanGlScott]

In case you need a CLC member 👍, this has my support.

[hvr]

I'll have take care of #19 as a pre-req first (which I intend to do soon); then we can go forward with this one.

[m-renaud]

I just submitted a GHC ticket (see here) which proposes changing the behaviour of -Werror to no longer fail compilation if a DEPRECATED symbol or module is used (unless explicitly requested with -Werror=deprecations).

This addresses one of the largest concerns I've heard raised about not bumping a major version when adding a DEPRECATED pragma, namely that because so many folks use -Werror, without a major version bump it would break a lot of builds.

I'm completely in favour of this policy change since a deprecation is only meant to inform users about planned changes in upcoming releases and shouldn't affect the buildability of the current major version. Of course, we'll need to wait for whatever version of GHC that gets released in (if accepted) to reach wide adoption, but its better than never fixing it!