These notes are from a discussion between Steve Jacobs and Mel Chua on 27 Sep 2019, in Atlanta GA. They are rough, raw braindumps; they may not be readable by others yet, and that's okay. All part of trying to do this work in the open as much as possible.
Information is available, not necessarily seen
Participation, usage, and roles (both upstream and downstream) are often self-initiated and not clearly recorded
- financial / operating costs
- servers, etc.
- people time
- bus factor; are there enough contributors, can people step away?
- is there interest in it existing? (incl. but not limited to users)
- people being involved, new people coming in, diverse types of people
- keeping up with responses to incoming issues in a timely manner
- processes/styles/etc. up-to-date with modern standards ("what a modern contributor would expect")
- test coverage
- issue tracker
- automated testing
- specifications for what it should do
- ability for people with basic skills to contribute
- tension re: corporate control
- pays for time spent, but "would do anyway, just slower and at cost to family"
- FOSS people often filter potential employers by whether they'll allow and not constrain FOSS work
- project management is the hard part to find!
- predates web frameworks
- backstory: previous python.org project failure taught PSF the need for project management
- "in flight, but never really got over the finish line"
- got server space donated; would be HUGELY expensive if not
- side project that became production
- "mudball architecture"
- this is terrible, let's start from scratch
- don't know what people are using how; what can we break, what shouldn't we?
- communications: you can try everything, but some people will still ignore it
- big win: people can contribute now, fed into the energy ("stickers")
- maybe we should turn it off, that's how we get people to pay attention
- "there's info out there, I just never kept up, it's too much"
- heard through twitter, friends, etc. "but I guess there's the blog and mailing list also"
- people-to-people connections important
- assumed it was magic
- "how did i not see that mess?"
- don't need to know, don't want to know; it's plumbing
- always worked, never thought about it (didn't evaluate it the way we would evaluate a library)
- "i feel weird that my company never knew python needed help"
- tension between "oh yeah, of course!" and "it's still magic"
- "Donald, the hero of PyPI - wait, needs a job?"
- we could have done the technical transition any time, it's the communication and financial part that was hard
- "breaking stuff on users" - infra upgrades are not like library upgrades because you have to decide when to switch the service over, whereas users decide when to transition between libraries
- "it's fun!"
- project management: less fun
- we don't make people sign up, track, etc. - pros and cons
- needed a way to handle money
- needed to learn to hire project managers
- needed a track record for applying for grants (wants a more open process later)
- could run PyPI as a business, has decided not to
- limited by international laws
- management of donations, volunteer relationships, etc. takes up paid staff time