Re: wither the Platform

On 2015-03-25 at 06:52:22 +0100, Mark Lentczner wrote:

On Tue, Mar 24, 2015 at 8:20 PM, Gershom B wrote:

...

install Yesod, or GHCJS, or Yesod and then GHCJS, and then some package with an API binding for some webservice which has not been updated in two years and requires an old version of time, and then maybe a GUI toolkit and of course lens.

That sounds like a recipe for Cabal Hell, Platform or not!

Regardless of the hellish issue, Gershom's comment indirectly highlights of one thing where I'm wondering if the HP's growth isn't bounded by diversity: There are some areas which I'd expected to some degree in a batteries-included platform, where the Haskell ecosystem has diverged into popular but distinct package-sub-ecosystems (which all have their respective communities/followers), such as HTTP-serving (Yesod/Snap/Happstack/...), or which lens-abstraction to use, or at the more fundamental level, even the streaming abstraction (pipes/conduit/io-streams/machines/...) doesn't seem to have a clearly recommended and agreed upon representative. Also, to this day we don't have any TLS library support in the platform, which also is subject to debate of which crypto-library to use (and there's also the question whether to use OpenSSL via FFI or a native TLS reimpl). So the platform-included `HTTP` package is not even able to access `https://` URLs which is quite sad, as this also holds back `cabal-install`'s ability to access `https://`-only repositories. So, where do you see the platform's growth for those packages/areas where you'll probably not get a reasonable majority consensus for picking a specific package? Cheers, hvr