Re: Proposal: Expanding the CLC

Absolutely. I think there could very well be a platform that focuses solely on authorship. This could be made easier by enforcing package names to contain the authors username and then relying on package-qualified imports or something. Just random thoughts. But yeah, hackage is not that. It already contains various means to deal with maintainership that exceeds authorship. There's a good counter-argument to my proposal though, namely that it may undermine a users trust in a package, which may be based on the authors quality standards and their attitude. I agree it is an argument, but I also believe a body like the CLC has sufficient competency to make up for this. And it's also a reminder to every maintainer: make sure you have co-maintainers you trust, such that this will never become a problem. On February 13, 2021 8:03:04 PM UTC, John Ericson wrote:

Yeah I strongly agree with the sentiments here, and the concrete measures. Thank you, Emily, for proposing them.

I assume some people will be worried about undermining the prerogative of individual maintainers. My view is this *not* a good reason to "hold

the CLC back". A bit off topic, but In the long term, I am optimistic for technical solutions to make dealing with libraries and versions, alternative ecosystems, etc. easier. Basically I want it all---both collaborative ownership and authorship, and healthy decentralized experimentation---and I think that's possible.

John

On 2/12/21 1:46 AM, Julian Ospald wrote:

...

Hi,

my opinion is that we should develop a zero tolerance towards unresponsive maintainership.

Contributors shouldn't have to escalate on the ML and shouldn't have to request package takeovers. These things are awkward and require more dedication than necessary to be a valuable co-maintainer.

The CLC should proactively scan for popular packages that require new

...

maintainer juice, contact the current maintainers and call for help on the ML (whether core/boot/something-else doesn't really matter to me... redefine the CLC competencies if you must).

I've had many PRs over the years that took 6-12 months for a response. This is not an acceptable response time.

Cheers, Julian

On February 11, 2021 11:54:19 PM UTC, Emily Pillmore wrote:

Hi All,

Over the past year, two things have become increasingly clear to me as I've carried out my CLC duties:

1. The CLC is under-resourced. This is evidenced by the fact that several maintainers who are not CLC members have been forced to step up to help take on some of the maintenance burden for many of the CLC libraries. Namely, `vector`, `bytestring`, `random`, `unix`, and more. The current CLC head count is not enough to dedicate at least one maintainer per package, which is leading to us all being spread thin, and the less-loved packages are falling into disrepair as a result. Couple this with the fact that roughly half the CLC do not have these packages actively within their maintenance cycles, and we arrive at the current problem.

2. The current set of "core" libraries does not cover what is generally considered "core" in the community. From now on, I'll refer to "core" packages as "boot" packages, and identify core packages to be those that are have proven to be incredibly popular tools for building things in Haskell. For example `zlib`, `parsec`, `regex-base`, `regex-posix`, `network`, etc. In particular, if any of these core packages saw their current authors disappear, or incapacitated in any sense, it would seriously harm the Haskell ecosystem. `cabal-install`, for example, requires several of those packages as upstream dependencies. Currently, we are dealing with this nightmare situation where work is stalled across many packages due to a particular set of maintainers being very difficult to reach, one of whom having disappeared completely for all maintenance intents and purposes.

Ergo, we have a problem. Thankfully, many people have stepped up showing renewed interest in maintaining such packages with the latest crop of CLC folks, and this poses an interesting opportunity.

My proposal is this:

1. We expand the CLC from 9 members to 22 members such that we have at least 1 CLC maintainer per boot package. There are a large number of fantastic candidates already available, who would be perfect for the role. In fact, many of the candidates whom we would ask are already maintaining these packages. In particular, Andrew Lelechenko, Simon Jakobi, Viktor Dukhovni, Dominic Steinitz, Alexey Khuedyakov are already serving within this role (and thank you for it!). Andreas Abel has also offered to help take on one of the core packages.

2. We consider a dedicated "Haskell Action Team" (name and idea courtesy of Carter Schonwald) to oversee packages in the Haskell github repo that can act as supplementary maintainers for many of the core packages contained therein. Currently, there are many in need of help. `zlib` comes to mind, which is currently blocking `bytestring-0.11` migration work due to having no available maintainer with the permissions to do a release. This, in turn, is stalling `cabal-install`. Short of taking over the package, we would have to ask for an emergency Hackage release if the neither maintainer shows up to do it in a reasonable time frame.

This is just one step towards helping ease the burden of maintenance of so-called core and boot packages. I hope you agree that this is a good idea, and if we get enough thumbs up, then Chessai and I will draw up the necessary changes to the CLC remit and we'll get started!

Cheers, Emily

_______________________________________________ Libraries mailing list Libraries@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries