[Haskell-cafe] base version numbers

Ivan Lazar Miljenovic ivan.miljenovic at gmail.com
Thu Jun 16 09:11:38 UTC 2016

On 16 June 2016 at 18:16, Tom Ellis
<tom-lists-haskell-cafe-2013 at jaguarpaw.co.uk> wrote:
> On Thu, Jun 16, 2016 at 06:09:51PM +1000, Ivan Lazar Miljenovic wrote:
>> On 16 June 2016 at 17:58, Tom Ellis
>> <tom-lists-haskell-cafe-2013 at jaguarpaw.co.uk> wrote:
>> > Since each version of GHC depends precisely on one version of base, would it
>> > not be more clear to name them after their version of GHC, rather than
>> > 4.x.y.z?
>> Except that:
>> * That would break the Package Versioning Policy and thus make it more
>> difficult to determine to what extent there are any potentially
>> breaking changes
> Why's that?
> When GHC x.y.z depends on base a.b.c.d, a and b only depend on x and y, and
> vice versa.  Seems like this condition implies that the PVP would be upheld.

For anyone else wanting to use base as a dependency for their code
(AFAIK, you have to use either base or one of the Haskell* packages).
Furthermore, we might have a situation come up where we have base-5.*
and have a similar situation to some of the GHC-6.* releases that
shipped with two versions of base.  It may also cause problems with
versions for pre-releases of GHC.

>> * Makes it more difficult for an alternate Haskell implementation to
>> use the Prelude
> I don't understand that.  Surely an alternate Haskell implementation cannot
> depend on base?  It's far too GHC specific.

There are lots of CPP hackery in there, but it has at least in the
past been also used with Hugs, NHC and I believe JHC and AHC.

>> * Makes it more difficult for a potential future where versions of
>> base are less tightly coupled to GHC and can thus be installed
>> separately (rather than needing a new version of GHC to get a new
>> feature from base)
> Is that really realistic?

Probably not.

Whilst in practice releases of base are coupled to releases of GHC,
this might not always be the case (and I think i recall point releases
of GHC in the past that didn't ship with a version of GHC).

Ivan Lazar Miljenovic
Ivan.Miljenovic at gmail.com

More information about the Haskell-Cafe mailing list