16:00:18 <cohosh> #startmeeting tor anti-censorship meeting
16:00:18 <MeetBot> Meeting started Thu Oct 22 16:00:18 2026 UTC.  The chair is cohosh. Information about MeetBot at https://wiki.debian.org/MeetBot.
16:00:18 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:00:21 <cohosh> hello
16:00:36 <dcf1> hello
16:00:40 <cohosh> here is our meeting pad: here is our meeting pad: https://pad.riseup.net/p/r.9574e996bb9c0266213d38b91b56c469
16:00:48 <onyinyang[mds]> Hihi!
16:01:35 <cohosh> let us know if you need a link to edit it
16:01:46 <cohosh> i'll give a moment to add topics
16:01:50 <unic3rn> hello
16:04:18 <cohosh> ok the first topic is about deploying a new version of the snowflake broker
16:04:33 <cohosh> https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowflake/-/work_items/40573
16:05:09 <cohosh> there is a new rate limiting feature that will turn away proxies if there was a proxy poll from their IP address too recently
16:05:50 <cohosh> this is one of the final pieces of the proxy fairness proposal: https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowflake/-/work_items/40507
16:06:15 <cohosh> i'm anticipating it will cause a big change to availability of proxies
16:06:44 <cohosh> because i've heard there are several proxy operators that run multiple proxy processes behind the same IP address
16:06:57 <cohosh> and by multiple i mean, 10s or 100s of proxy processes
16:07:20 <cohosh> this means that the earlier changes to let the broker inform proxies how often to poll won't have made much of an impact on the pool, but this change might
16:08:09 <cohosh> so i want to be careful and i'm hoping to pick a date that at least other snowflake devs know about and perhaps can help monitor, adjust the poll rate, or even roll back the change if it's needed
16:08:41 <cohosh> we're a bit low capacity at the moment but we've got a lot of broker changes piling up so i'm thinking of giving it a try early next week
16:09:48 <cohosh> dcf1 you might be the only other snowflake dev here, do you have any opinions on this?
16:11:08 <dcf1> if you're unsure, one thing you can do is a pilot deployment with a planned revert
16:11:42 <dcf1> Like, choose a time, make the change, let it run for 2 hours, then revert the change no matter what and look at the results of what happened.
16:12:06 <cohosh> oh that's a good idea
16:12:49 <cohosh> i don't see a downside to it
16:13:01 <dcf1> I would say, feel free to pick a date and time that is convenient, and a temporary deployment like that has low risk.
16:13:20 <cohosh> ok i'll add that to the deployment plan, thanks!
16:13:24 <dcf1> The downside is just more fussing about and aggravation, versus just going ahead and doing it.
16:14:00 <dcf1> The point of doing a temporary trial deployment would be to check for showstopper problems, like if the matching rate suddenly drops to 1% or something.
16:15:23 <cohosh> yeah i think it's worth it for this change, for a two hour deployment i'll be able to learn enough from the prometheus metrics about how well it's working
16:16:34 <cohosh> okay, anything else to discuss today?
16:18:20 <cohosh> alright i'll end the meeting, thanks for joining :)
16:18:27 <onyinyang[mds]> thanks!
16:18:37 <cohosh> #endmeeting