The Cat Model of Package Ownership

Debian has been moving away from strong ownership of packages by package maintainers and towards encouraging group maintainership, for very good reasons: single maintainers have a bad bus factor and a number of other disadvantages.

When single maintainership is changed into maintainership by a small[1], open group of people who can easily communicate and sync with each other, everything is just better: there is an easy way to gradually replace people who want to leave, but there is also no duplication of efforts (because communication is easy), there are means to always have somebody available for emergency work and generally package quality can only gain from it.

[1]how small is small depends a lot on the size of the package, the amount of work it requires, how easy it is to parallelize it and how good are the people involved at communicating, so it would be quite hard to put a precise number here.

Unfortunately, having such group of maintainers for every package would require more people than are available and willing to work on it, and while I think it's worth doing efforts to have big and important packages managed that way, it may not be so for the myriad of small ones that make up the long tail of a distribution.

Many of those packages may end up being maintained in a big team such as the language-based ones, which is probably better than remaining with a single maintainer, but can lead to some problems.

My experience with the old OpenEmbedded, back when it was still using monotone instead of git[2] and everybody was maintaining everything, however, leads me to think that this model has a big danger of turning into nobody maintains anything, because when something needs to be done everybody is thinking that somebody else will do it.

[2]I've moved away from it because the boards I was using could run plain Debian, but I've heard that after the move to git there have been a number of workflow changes (of which I've seen the start) and everything now works much better.

As a way to prevent that, I have been thinking in the general direction of a Cat Model of Package Ownership, which may or may not be a way to prevent some risks of both personal maintainership and big teams.

The basic idea is that the “my” in “my packages” is not the “my” in “my toys”, but the “my” in “my Cat, to whom I am a servant”.

As in the case of a cat, if my package needs a visit to the vet, it's my duty to do so. Other people may point me to the need of such a visit, e.g. by telling me that they have seen the cat leaving unhealty stools, that there is a bug in the package, or even that upstream released a new version a week ago, did you notice?, but the actual putting the package in a cat carrier and bringing it to the vet falls on me.

Whether you're allowed to play with or pet the cat is her decision, not mine, and giving her food or doing changes to the package is usually fine, but please ask first: a few cats have medical issues that require a special diet.

And like cats, sometimes the cat may decide that I'm not doing a good enough job of serving her, and move away to another maintainer; just remember that there is a difference between a lost cat who wants to go back to her old home and a cat that is looking for a new one. When in doubt, packages usually wear a collar with contact informations, trying to ping those is probably a good idea.

This is mostly a summer afternoon idea and will probably require some refinement, but I think that the basic idea can have some value. Comments are appreciated on the federated social networks where this post is being published (friendica, pump.io), via email (valid addresses are on my website and on my GPG key) or with a post on a blog that appears on planet debian.

Send a comment: unless requested otherwise I may add it, or some extract, to this page.

Return to Top