The Future of community.haskell.org

Community.haskell.org is a server in our ecosystem that comparatively few know about these days. It actually was, to my knowledge, a key part of how the whole haskell.org community infrastructure got set up way back when. The sparse homepage still even says: "This server is run by a mysterious group of Haskell hackers who do not wish to be known as a Cabal, and is funded from money earned by haskell.org mentors in the Google Summer-of-Code programme." At a certain point after this server was created, it ceased to be run by a "mysterious group of Haskell hackers" and instead became managed officially by the haskell.org Committee that we know today. You can see the original announcement email in the archives.

The community server, first set up in 2007 played a key role back before the current set of cloud-based services we know today was around. It provided a shared host which could provide many of the services a software project needs -- VCS hosting, public webspace for documentation, issue trackers, mailing lists, and soforth.

Today, the server is somewhat of a relic of another time. People prefer to host projects in places like github, bitbucket, or darcs hub. Issue trackers likewise tend to be associated with these hosts, and there are other free, hosted issue trackers around as well. When folks want a mailing list, they tend to reach for google groups.

Meanwhile, managing a big box full of shell account has become a much more difficult, riskier proposition. Every shell account is a security vulnerability waiting to happen, and there are more adversarial "scriptkiddie" hackers than ever looking to claim new boxes to spam and otherwise operate from.

Managing a mailman installation is likewise more difficult. There are more spammers out there, with better methods, and unmaintained lists quickly can turn into ghost towns filled with pages of linkspam and nothing but. The same sad fate falls on unmaintained tracs.

As a whole, the internet is a more adversarial world for small, self-hosted services, especially those whose domain names have some "google juice". We think it would be good to, to the extent possible, get out of the business of doing this sort of hosting. And indeed, very few individuals tend to request accounts, since there are now so many nicer, better ways of getting the benefits that community.haskell.org once was rare in providing.

So what next? Well, we want to "end of life" most of community.haskell.org, but in as painless a way as possible. This means finding what few tracs, if any, are still active, and helping their owners migrate. Similarly for mailing lists. Of course we will find a way to continue to host their archives for historical purposes.

Similarly, we will attempt to keep source repositories accessible for historical purposes, but would very much like to encourage owners to move to more well supported code hosting. One purpose that, until recently, was hard to serve elsewhere was in hosting of private darcs repositories with shared access -- such as academics might use to collaborate on a work in project. However, that capability is now also provided on http://hub.darcs.net. At this point, we can't think of anything in this regard that is not better provided elsewhere -- but if you can, please let us know.

On webspace, it may be the case that a little more leniency is in order. For one, it is possible to provide restricted accounts that are able to control web-accessible files but have no other rights. For another, while many open source projects now host documentation through github pages or similar, and there are likewise many services for personal home pages, nonetheless it seems a nice thing to allow projects to host their resources on a system that is not under the control of a for-profit third party that, ultimately is responsible to its bottom line and not its users.

But all this is open for discussion! Community.haskell.org was put together to serve the open source community of Haskell developers, and its direction needs to be determined based on feedback regarding current needs. What do you think? What would you like to see continued to be provided? What do you feel is less important? Are there other good hosted services that should be mentioned as alternatives?

And, of course, are you interested in rolling up your sleeves to help with any of the changes discussed? This could mean simply helping out with sorting out the mailman and trac situation, inventorying the active elements and collaborating with their owners. Or, it could mean a more sustained technical involvement. Whatever you have to offer, we will likely have a use for it. As always, you can email admin@h.o or hop on the #haskell-infrastructure freenode channel to get involved directly.

Written by gershomb on Feb 23 2015, 12:43 PM.
busybody
Projects
None
Subscribers
None