Ticket #107 (closed enhancement: fixed)
ijones suggestions for HAC07
|Reported by:||ijones||Owned by:|
Description (last modified by ijones) (diff)
Here is a list of suggestions from Isaac about what needs to happen to get Hackage off the ground. Above all, though, you hackers should make it your own!
Isaac proposes two high-level goals. See details below:
- "Finish" Hackage, version 1.0
- Make a proposal to Isaac / Galois about the server needs for Hackage. We cannot give too many people root or login accounts on haskell.galois.com; There is too much mission-critical software there. At the same time, we'd like to support many smaller projects; every few days someone asks for someplace to host their darcs repo!
- should the Hackage system be both a package database AND a (darcs?) hosting site for Haskell-related projects?
- who should have root access to this machine (might inform later questions)
- who should have user accounts on this machine
- should it live on Monk (with the same set of sysadmins that exist now)
- should we get a new machine and / or use a virtual colocated machine ( http://vcolo.com or somesuch). we could add more root users this way
- what should the machine be used for?
- host projects with darcs
- host hackage
- host gforge
- host trac
- host community blogs, etc
- what should the machine be called
- hackage: migrate current hackage.haskell.org stuff?
- darcs: migrate current darcs stuff?
- community.haskell.org: the machine for the people
- haskellforge: or whatever.
over-all tasks to improve adoption
It's great to write good code, but that doesn't mean people will use that code! How do we get Hackage adopted? (See also Isaac's packages-6.6 blog entry.)
- create and maintain precise instructions for exactly what to do to find and install packages.
- simply upload as many cabal packages as you can find
- use it ourselves, note problems, fix them.
- and make sure that you can actually install them with cabal-install
- once you have a set of packages that can be installed, works with a known version of cabal (and GHC?) move them to "testing" as described here.
Cabal-install is currently living in a darcs branch as described in CabalWithInstall?. Its basic functionality is good.
Some tickets that may or may not be up to date.
- integrate the cabal-with-install branch into the mainline Cabal source tree
- fix up the makefile so that "sudo make install" installs both the cabal library and cabal-install
- what about cabal-setup? does cabal-install currently use it? if so, we should install that as well.
- play with cabal-install, discover its shortcommings and add tickets to this trac
- and remove any out-of-date tickets that you find.
- user interface tweaks. does it work well? is it smooth?
- cabal-install install cabal-install should smoothly upgrade itself :)
- pick a very complex package with lots of dependencies and make sure cabal-install works for it.
- can hackage itself be installed with cabal-install?
Cabal-put is currently a script living on the hackage.haskell.org server. This is nice because people we already know & trust can upload stuff into hackage w/ cabal-put.
- make a web front-end for uploading? (Ross P. has considered some security ideas)
- remember who uploaded, make them the maintainer? or have the idea of a hackage maintainer separate from the upstream maintainer?
The collection of packages lives here. Debate topic: Should Hackage be just the database of packages and the UI (which would be great) or should it expand to be a (darcs?) hosting platform for projects?
- front-end for viewing, searching, querying the package repository. Lemmih has done good work here, check with him! His front end does work with the new repository layout!
- integration with darcs?
- integration with trac? (each new package is a component?)