Agenda 5/12/2015

crates.io and downtime (acrichto)
meetings, subteams, oh my (nmatsakis)
Servo rustup currently blocked on weird value init issue, will try to get repro (larsberg)

https://etherpad.mozilla.org/Whistler-Servo-Training, https://etherpad.mozilla.org/Whistler-Servo-Meetings

Servo

larsberg: working on servo rustup. some of our static thread locals are no longer being initialized. super weird. will get a repro. everybody is exhausted from last rustup so we're going slowly.
larsberg: stuff happening in whistler: training: (missed). I'll welcome feedback and will be asking for assistance. In the middle of arranging rooms and talking to managers.

larsberg: other etherpad is meetings between servo team and other teams at mocos. Let me know if you want other meetings with other teams or need to change agendas; I'm marshalling meetings and rooms already anyway.

steveklabnik: several gecko people have been on IRC lately
larsberg: basic rustc support landed in gecko, mostly driven by people outside research
nmatsakis: re training, do you need people to develop content? how many people?
larsberg: i'm getting a lot more interest than expected. was hoping to get a room for 60, but may not be big enough. 3 hours shouldn't be too bad given felix's materials from icfp.
larsberg: open to any feedback about how to put this together. been trying to get feedback from the gecko teams about what would be useful to them.
nmatsakis: steve's done this recently
steveklabnik: done several from 1 - 5 hours. this will be easier because your audience knows c++.
aturon: making a list of people who want to be involved would be a good step to make sure it doesn't fall through the cracks on our end
pnkfelix: i want to help
aturon: and i
larsberg: (swoons)
nmatsakis: (waves silently to indicate his interest)

acrichto: crates.io went down Sunday. I didn't hear about it until hours later. All I had to do was restart. Underlying problem is architectural and hard to fix.
nmatsakis: does it make sense to automatically restart?
acrichto: if i knew how to do that. restarting only requires heroku push privelege. we can expand who has that.
brson: how bad is it going to crash friday?
acrichto: it's tied to github. if github pushes take a long time. the server itself won't go down. i'll try to watch closely over the weekend
erickt: one simple tool that might help - monit (https://mmonit.com/monit/)
acrichto: I've heard of it, but this is on heroku so (...)
erickt: you could script something with monit to automatically restart
brson: what does it do?
erickt: it watches the health on various metrics. so it could e.g. hit a url every few minutes and do something if its down
brson: problem is that crates.io goes down when github goes down. just restarting isn't guaranteed to fix anything?
acricho: yeah, but once the server is restarted it frees up all the threads again and it's good for a bit, but yeah if github is down for awhile crates.io may have to be restarted more than once
acrichto: not pressing enough to deal with right now

(missing talk about crates.io dns problems)

acrichto: name servers are hosted by zerogo (?), hosted by heroku. every so often crates.io just stops resolving in some parts of the world. not clear why. this morning i had errors this morning in honduras. brian had errors this morning.
erickt: maybe they're just caching the names and pointing at old heroku servers
nrc: similarly, what's the offline story with cargo. one reason this was causing problems is that cargo stops working. i have a student working on refactoring and he can't use rust on their university machines. he's not even using crates and he can't because of their firewall.
acrichto: cargo should never touch the network unless

(brson complaining about using cargo offline)

erickt: would be nice if, if it fails to connect to the world it's not a fatal error
larsberg: I think we've opened bugs on most of the ones we've found on servo. Only places were we have weird behavior is around wildcard deps and path overrides. I suspect we're off the beaten path on some of those.
acrichto: somebody reduced the path dependency bug. case erickt and brian are talking about with the registry update is possible.

nmatsakis: moving toward implementing subteams. Two questions: 1) how much longer to have this meeting and what to replace it with; 2) how to have face-to-face contact while having more open meetings
pnkfelix: when will we know who the subteams are?
nmatsakis: I think most people have responded. I think we'll announce that soon.
brson: Do we all know who is on the subteams?
nrc: Is aaron running the libs team? Niko language design?
nmatsakis: yes
nrc: Do you know who is going to be on the libs team?
aturon: [redacted people]
nmatsakis: i'd like to have subteam specific meetings because one appealing thing to me is to have deep-dives with a focused audience. but I think it's better to do one-off topic-specific meetings.
pnkfelix: one thing with standing meetings is that you have to know to keep that block of time free
aturon: trying to fit more volunteers in makes standing meetings hard
nmatsakis: could push for more discuss-based communication but that's so different
nrc: wonder if for volunteers its more or less hard to have a reserved slot. busy people might have an easier time reserving a time
pnkfelix: different subteams may have different policies
acrichto: not entire team needs to make meetings
nmatsakis: in the compiler there are people who focus on different aspects
aturon: draft document about how subteams might operate suggests not to use meetings for decision making. give volunteers more capacity to get involved.

(lots of discussion about precisely the correct volume of meetings and how to involve volunteers)