[Nix-dev] Typing nix − funding campaign

[Théophane Hufschmitt]

Thu 30 Mar 17 − 00:23, Jeaye(contact at jeaye.com) a écrit:
> Congrats on the funding!

Thanks :)

> As someone who's been looking for more static validation of my Nix expressions, this is an exciting change. I must admit, however, that I wonder if Nix is just not the "right" tool for the job. I didn't see anyone else in this thread mention it, to my surprise, so I'll bite: if you have some funding and some time, why not integrate an existing language, with a working, mature, static type system instead?
> 
> This is where I think Guix made a better choice: many people already know Scheme and are comfortable with it. If Haskell, for example, has the kind of type system we're seeking, what benefit do we get for sticking with Nix? Only the sunken cost?
> 
> If it isn't clear, I think that purely functional package management is a wonderful thing, but I don't think a new (Nix) language is the best way to use it.

I don't think a dedicated (and new) language is a bad think because
nixpkgs is a quite hudge codebase (and could be much bigger) where most
of the code follows exactly the same few patterns (calls to
`mkDerivation` or nixos module declarations), so I think the cost of
maintaining a language optimized for those patterns is worth it (for
scheme, lisp users tend to think that it is flexible enough to fit as a
perfect DSL for any use case, so if this is indeed true, it also is an
excellent choice − I always wanted to try some lisp but never found the
strenght to fight those cohorts of parens so I couldn't make my own
opinion on this).
And the real cost of switching to a new language would probably be −
much more than the sinking cost − the unavoidable split of the community
(which already isn't that big...).


(And for haskell in particular − at least GHC-haskell −, I really
wouldn't like it for nixpkgs because I *really* wouln't want nix to
depend on something as huge as GHC).

That being said, I agree that nix is somehow weird (and I agree even
more after having worked a bit on the grammar), and I would have by far
preferred something closer to classical functional languages.

-- 
Théophane Hufschmitt
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.science.uu.nl/pipermail/nix-dev/attachments/20170330/2a7d8a6c/attachment.sig>