21:59:02 <pastly> #startmeeting Simple Bandwidth Scanner (sbws) Meeting
21:59:02 <MeetBot> Meeting started Thu Apr 19 21:59:02 2018 UTC.  The chair is pastly. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:59:02 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
21:59:07 <pastly> Pad: https://pad.riseup.net/p/BenwGdgpz0uS
21:59:23 <pastly> Hello people interested in sbws.
22:00:09 <juga> hi pastly
22:00:19 <juga> why sbws
22:00:33 <juga> not bandwidth scanner in general?
22:01:08 <pastly> Ah. Is there non-sbws bw scanning stuff to talk about?
22:03:05 <pastly> (Just want to briefly thank ln5 again for the server space to help make sbws a real thing)
22:03:50 <pastly> teor4: are you able to make this meeting?
22:05:21 <teor4> I am here
22:05:47 <pastly> Well. I left some updates in the pad. Big things I did last week are mostly a painful logging library switch and starting a dirauth in the testnet.
22:06:43 <pastly> juga helped me remember that someday soon I'll hopefully have time during the day to work on a peerflow-like thing, so maybe I should be making sure sbws is headed towards completion
22:07:22 <pastly> Which is good, because I think it is nearly ready.
22:07:38 <pastly> Nothing big off the top of my head needs implementing.
22:07:56 <pastly> docs, tests, and user support are the primary things I want on my plate :)
22:08:20 <pastly> anybody else want to mention some updates?
22:08:33 <pastly> Regarding sbws or other things?
22:09:49 <juga> pastly: in pad
22:10:31 <juga> hmm
22:10:37 <pastly> "once sbws repo is public, can we work on new releases publicly?"
22:10:40 <pastly> Yes.
22:10:44 <pastly> Good question
22:10:54 <juga> pastly: thanks for the answer
22:11:44 <pastly> I'm thinking I could quickly throw the sbws docs on an onion service in the next ~day. Would that be helpful to anyone?
22:11:58 <pastly> (and initially only share it privately)
22:11:59 <juga> sure
22:12:18 <juga> what about having them in rtfd if repo public next week?
22:12:57 <pastly> As soon as the repo is public I plan on looking at putting them on readthedocs.
22:13:07 <juga> k, thx
22:13:10 <pastly> I just don't want to have to give them my password or something.
22:13:16 <pastly> I'm sure that's not required for a public repo
22:13:55 <juga> u'll need to give them access your gh via oauth
22:14:27 <pastly> alrighty
22:15:41 <pastly> While talking to a dirauth, they expressed concern about trusting people other than themself to run sbws servers.
22:16:23 <pastly> And that they may not have a good enough machine handy to run a full helper themself.
22:16:58 <pastly> So even if we can convince some relay ops to run sbws, I'm not sure many of the dirauths will want to use "random" helpers.
22:17:43 <pastly> I see a few ways forward other than just ignoring this and hoping it works out okay.
22:18:10 <pastly> 1. completely switch fundamentals and use HTTP(S). Then instead of sbws servers we'd have web servers, potentially on CDNs.
22:19:06 <pastly> 2. Run big public-ish sbws servers that aren't near relays. Add encryption to the wire protocol so auth passwords aren't leaked. And don't require peopel to run sbws servers
22:19:24 <pastly> And that's the end of my ideas.
22:20:16 <teor4> You could ask the dirauth list which option they prefer
22:20:18 <juga> 2. could still be close to helper relays
22:20:18 <pastly> Anything sound significantly less bad than the others?
22:20:45 <teor4> 1 is a lot less effort to maintain in the long term
22:20:52 <pastly> teor4: what's the list name?
22:21:41 <teor4> see pm
22:21:48 <pastly> ln5: unless I'm greatly mistaken, you run a dirauth. Any of these options sound not bad? Would you run a bwauth with any of these?
22:21:51 <pastly> teor4: thanks
22:22:20 <pastly> juga: yeah, they could. Could run some in the biggest data centers for tor relays
22:23:55 <ln5> pastly: what does "a full helper" mean? a non-exit relay with no bw cap?
22:24:12 <pastly> teor4: I really appreciate the input you've been giving on sbws tickets. Do you think network team involvement will pick up once sbws is open?
22:24:20 <teor4> another advantage of 1 is that it can re-use the existing bandwidth servers
22:24:51 <pastly> ln5: the sbws server + a relay that (ideally) is underutilized and capable of a lot of bw so it isn't a limiting factor in measurements
22:25:04 <juga> teor4:
22:25:12 <teor4> pastly: you could ask isis or catalyst to be involved once the code is open
22:25:17 <juga> ops
22:25:24 <teor4> they are assigned to the task on the roadmap
22:25:57 <ln5> i don't see how this is so much different than the current situation where ppl either run their own bandwidth server (a web server) or use the one from the setup docs
22:26:28 <ln5> or sorry, i see how the resources required are higher
22:26:36 <pastly> yeah. crypto mainly
22:26:43 <pastly> And no CDN option
22:26:53 <ln5> but for the question of trusting someone to run one, i mean
22:27:30 <ln5> i am running a dirauth but i might not be representative since i take bw data from someone else, whom i trust
22:27:58 <pastly> Alright. Thanks
22:28:20 <ln5> i'd be happy to try running sbws though, fwiw
22:28:47 * pastly wonders if we could convince Tor to run a few scattered sbws servers
22:29:08 <pastly> (for some definition of "Tor" ...)
22:29:47 <pastly> ln5: maybe you can be the first real dirauth to try it :)
22:30:42 <pastly> Ok so I'll ask the dirauth list what they think regarding this
22:30:47 <teor4> I think the scaling will help with some of the resource issues
22:30:47 <juga> teor4: any thoughts about the tasks?
22:30:47 <ln5> Tor Project Inc doesn't run any parts of the tor network but there are other organisations who are pretty trusted that do
22:31:10 <teor4> so if you don't have enough CPU, you can scan the network slower
22:31:49 <teor4> and if you can't sustain high speeds due to CPU, you will scale all your results, but the order will still be correct
22:32:34 <pastly> Unless $large_number relays all have essentially the same result since they can do more than you're measurement setup can measure
22:34:33 <teor4> that's fine, they will still be scaled large
22:36:07 <teor4> people can always stay on torflow
22:36:30 <pastly> Hehe yeah. Can't be that bad if we're still using it :p
22:36:34 <pastly> Is there anything else we should talk about? Any questions? Concerns?
22:37:20 <pastly> Everyone know what they want to work on, whether its sbws-related or not? :p
22:38:16 <juga> pastly: maybe we can have some informal follow up around 10amUTC somewhere?
22:38:23 <juga> or on monday
22:38:38 <pastly> juga: sure.
22:38:49 <pastly> 1000 utc is fine
22:38:59 <juga> just cause i'm toooo sleepy, and partly think missing teor4
22:39:45 <pastly> I'm excited that we seem to be making big, real strides to replacing torflow
22:40:27 <juga> pastly: that was one of the ideas :)
22:40:38 <teor4> it's also important to have a process that works for any new scanner
22:41:17 <pastly> Yeah.
22:41:38 <pastly> This might be useful for whatever I/we at NRL come up with next.
22:42:16 <pastly> And I don't think I need to be taking up any more of y'alls time this morning/evening/whatever.
22:42:44 <teor4> ok, thanks
22:42:54 <pastly> Thank you all for coming. Thanks for your support.
22:43:04 <juga> thx pastly, sry for my sleepiness
22:43:17 <pastly> Catch you all on the flippity floppity ;)
22:43:23 <pastly> juga get some sleep :p
22:43:25 <pastly> #endmeeting