[Haskell-cafe] Safe Haskell?

David Feuer david.feuer at gmail.com
Sat Apr 17 15:56:28 UTC 2021


There's really only one situation where I've personally felt I needed
something like Safe Haskell—and Safe Haskell doesn't deliver it.
Specifically, it's very hard to know whether it's safe to use a particular
class method at a particular type.

On Fri, Apr 16, 2021, 3:03 PM Richard Eisenberg <rae at richarde.dev> wrote:

> Hi café,
>
> Do you use Safe Haskell? Do you know someone who does? If you do, which of
> Safe Haskell's guarantees do you rely on?
>
> Here, a user of Safe Haskell is someone who relies on any guarantees that
> Safe Haskell provides, not someone who makes sure to have the right
> pragmas, etc., in your library so that users can import it Safely.
>
> Context: Safe Haskell is not lightweight to support within GHC and the
> ecosystem. Despite being a formidable research project with a (in my
> opinion) quite worthwhile goal, it's unclear which of Safe Haskell's
> purported guarantees are actually guaranteed by GHC. (The lack of
> unsafeCoerce is not actually guaranteed:
> https://gitlab.haskell.org/ghc/ghc/-/issues/9562.) Recent design
> questions about what should be Safe and what shouldn't be (somehow cannot
> find the discussion after a few minutes of searching; perhaps fill this in)
> have been answered only by stabs in the dark. The status quo is causing
> pain: https://gitlab.haskell.org/ghc/ghc/-/issues/19590. There are
> hundreds (maybe thousands) of lines of delicate logic within GHC to support
> Safe Haskell. These parts of GHC have to be read, understood, and
> maintained by people with limited time.
>
> I thus wonder about deprecating and eventually removing Safe Haskell. I
> don't have a concrete plan for how to do this yet, but I'm confident we
> could come up with a migration strategy.
>
> The set of people who would win by removing Safe Haskell is easy enough to
> discover. But this email is intended to discover who would be harmed by
> doing so. If you know, speak up. Otherwise, I expect I will write up a GHC
> proposal to remove the feature.
>
> Thanks,
> Richard
> _______________________________________________
> Haskell-Cafe mailing list
> To (un)subscribe, modify options or view archives go to:
> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> Only members subscribed via the mailman list are allowed to post.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20210417/ca72e771/attachment.html>


More information about the Haskell-Cafe mailing list