[Haskell-cafe] Package takeover (also: the future of Netwire)

[Ertugrul Söylemez]

Hi everybody,

I'd like to take over the following Hackage packages with my account
named [esz]:

  * acme-schoenfinkel,
  * cascading,
  * continue,
  * contstuff-monads-tf,
  * contstuff-transformers,
  * contstuff,
  * dnscache,
  * fastirc,
  * ihttp,
  * instinct,
  * ismtp,
  * netlines,
  * netwire,
  * quickset,
  * web-page,
  * webwire,
  * yesod-tableview.

These are actually already mine, but because my old account has become
practically inaccessible, I'm now following the official [takeover
procedure], asking myself whether I'm okay with that. =)

After the takeover, I'm going to deprecate the following packages,
unless someone would like to take over maintenance:

  * cascading: Clay is more comprehensive and easier to use.

  * contstuff*: Childhood experiments, because after writing a
    monad tutorial of course you must write an mtl replacement.

  * dnscache: Useful library, but needs maintenance.

  * fastirc: irc is reasonably efficient now (it used to be a String
    parser).

  * ihttp: use http-client.

  * ismtp: Useful library, but needs maintenance and should be rewritten
    using a modern streaming library.

  * netlines: use a modern streaming library.

  * netwire: Useful library, but needs maintenance and has high
    maintenance cost due to its design.  I'll likely replace it by a new
    abstraction at some point, using a different package name.
    Currently I recommend using one of the following libraries for FRP:

      * reactive-banana: My former favourite when it comes to push/pull
        FRP.  Well designed, clean semantics, reasonably efficient.

      * reflex: My current favourite when it comes to push/pull FRP.
        Well designed, clean semantics, reasonably efficient.

      * Yampa: AFRP for games, simulations and other real-time
        applications (for when a high and predictable framerate is more
        important than pushed events, same target domain as Netwire).

  * webwire: Experiment with FRP for handling server-side sessions in
    web applications.  Works with some caveats, not that useful in
    practice.

  * yesod-tableview: needs maintenance, not that useful in modern web
    applications.

That leaves the following packages for active maintenance by me, which
includes making sure they compile, keeping them free of bugs and
verifying/applying contributions.  Most of these need maintenance right
now:

  * acme-schoenfinkel: Surprisingly useful joke package.  Needs a few
    adjustments to compile with modern GHC.

  * continue: Monad transformer for named reentry points, isomorphic to
    CofreeT.

    Ideally this should be just an interface to CofreeT from 'free', but
    unfortunately that one's instances are more restrictive than
    necessary (Alternative instead of Plus), making it useless.  You
    couldn't use reasonable data structures like HashMap or Map.

  * instinct: Efficient feed-forward neural networks with backprop
    learning.  To be replaced by a more general machine learning
    framework at some point, but still useful on its own.  Needs
    maintenance.

  * quickset: Compact query-only binary search arrays.  Needs a more
    comprehensive API.

  * web-page: Component system for web pages.  Needs a more general API,
    because currently it makes too many assumptions on which packages
    you use (blaze-html, clay and jmacro).  Note: do not confuse with
    Athan Clark's 'webpage' library.


[esz]:                https://hackage.haskell.org/user/esz
[takeover procedure]: https://wiki.haskell.org/Taking_over_a_package


Greets,
Ertugrul
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 472 bytes
Desc: not available
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20160614/7391f8ee/attachment.sig>