17:00:41 <nickm> #startmeeting weekly net team meeting !1!1!
17:00:41 <MeetBot> Meeting started Mon May  9 17:00:41 2016 UTC.  The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:41 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:00:45 <nickm> hi all!
17:00:48 <teor> hi
17:00:54 <nickm> I'm in a room with various folks in montreal
17:01:03 <armadev> hello world!
17:01:10 <nickm> last week I did lots of ticket-wrangling, review, and merging
17:01:26 <nickm> working on review with gitlab was lovely. I think any code rview tool online would be lovely too, especially for big patches
17:01:39 <asn> hello!!!!
17:01:40 <isis> hey hey o/
17:01:41 <nickm> This week I'm in montreal talking about hidden services and hacking on them with folks.
17:01:42 <isabela> o/
17:01:42 <dgoulet> hello
17:01:46 <nickm> Maybe I'll find time to get other tickets done too.
17:01:51 <nickm> who's next?
17:01:55 <teor> I can go
17:02:02 <armadev> last week i looked at proposal 224; thought more about auth cert download schedules; and thought more about consensus download timing too. (done)
17:02:05 * sysrqb lurks
17:02:39 <teor> Last week I wrote a log message for #13953
17:03:00 <crazy_qubit> hi
17:03:04 <teor> which reminds operators there's only one IPv4 address in a relay descriptor
17:03:31 <nickm> yawning, athena, ping, if you happen to be around.
17:03:48 <athena> i've just posted a code review for #18809; got another one on the way for #18616
17:03:50 <teor> Also #18963, which remembers which directory gave us a consensus or certificate, and tries it first for certificates
17:03:59 <athena> (pong)
17:04:06 <asn> (i can go next)
17:04:13 <special> (also ready)
17:04:34 <Yawning> hm
17:04:35 <asn> Hello. During last week I did some gsoc, some code review (#17101), merged some
17:04:36 <teor> Also #18816
17:04:37 <Yawning> hi
17:04:38 <asn> prop224 spec changes, and also looked at the entry guard (prop259) code from
17:04:40 <asn> olabini and team. This week I will be focusing on prop224 stuff! Done. Next?
17:05:15 <teor> also #18982
17:05:23 <nickm> remember to use # item tags for discussion topics.
17:05:23 <Yawning> poked at obfs5, fixed the sha3 code
17:05:46 <special> I did some more finishing work on the 224 descriptor decoding branch and did some improvements to dgoulet's encoding branch. We'll be doing lots of review of those this week. I'm also about to push a revision of #15588 to maybe finally get that merged.
17:05:50 <Yawning> I could probably clean up/improve our sha3 code further, but what's there should be sufficient till openssl has propper support for it
17:06:04 <teor> also #18483
17:06:09 <isabela> Yawning: can we have a trac ticket for obfs5 - just so I can count it when doing reports etc
17:06:15 <Yawning> uh
17:06:17 <teor> oh and #18929
17:06:24 <Yawning> I guess?
17:06:39 <teor> and #18921
17:07:10 <Yawning> I'm currently trying to figure out how I want to do link padding
17:07:14 <Yawning> and how clever I want to make it
17:07:37 <Yawning> with the biggest stumbling block being "the clever algorithms require shit that probably won't work on OSX relaibly"
17:08:26 <Yawning> if someone knows know to get pmtu information on a per connection basis on osx, from userland let me know
17:08:48 <Yawning> (no, TCP_INFO does not do it, at least not on freebsd, which I assume they rip off)
17:09:53 <Yawning> I'll probably end up either shipping a "not as clever" algorithm with profanity and snark directed at OSX in the comments, or limiting the better algorithms to platforms I have (So, Linux)
17:10:13 <isis> last week i did some some small fixes to bridgedb #18237 #18949
17:10:17 <isis> i also did my OTF reports… still need to do the april one
17:10:21 <isis> then i got word that the FBI was going to subpoena me, so i published a blog post i had already written
17:10:23 <nickm> for this week, Yawning, I'd like us to really figure out the breakdown of tor-vs-pt tasks for your contract, and get it all nailed down.  I think it should be feasible if focus.
17:10:25 <isis> since then i have been dealing with switching lawyers (the EFF is representing me now)
17:10:28 <nickm> Yawning/ isabela : agree?
17:10:29 <isis> i am about to do some interviews with CNN and the Intercept and then get back to real work
17:10:33 <isis> also i finished and sent in the PQ handshake proposal and incorporated feedback
17:10:39 <Yawning> nickm: that'll be nice
17:10:40 <athena> Yawning: i'd recommend having a not-so-clever but portable fallback and then a real algorithm when it works anyway
17:10:47 * isis is still bad at copy/paste in qubes…
17:10:47 <Yawning> having a contract would be nice
17:10:57 <Yawning> athena: yah, was gonna use the obfs4 one as baseline
17:10:58 <nickm> isis: all of us here in montral are glad to hear about EFF lawyers.
17:11:12 <Yawning> cuz the better ones guzzle bandwidth like popcorn
17:11:18 <athena> remember that pmtu discovery is outright broken on a significant number of networks because of cargo-cult security on routers blocking ICMP can't fragment messages
17:11:23 <Yawning> yeah
17:11:29 <Yawning> it's also totally fucking broken on tails
17:11:34 <Yawning> for the same reason
17:11:40 <armadev> athena: thanks for the 18809 review!
17:12:05 <isis> nickm: yeah, i am pretty grateful to the EFF for their help :)
17:13:14 <dgoulet> isis: o/ good news
17:13:43 <Yawning> sinc eobfs5 starts off writing a few k of data, broken pmtu prolly results in worse behavior
17:13:45 <Yawning> oh well
17:13:50 <isis> if i missed any feedback in the tor-dev thread, someone please yell at me
17:14:13 <nickm> (I had to learn what SIDH was)
17:14:17 <Yawning> isis: think you got most of mine
17:14:22 <nickm> anybody who hasn't done a status-dump yet?
17:14:23 <isabela> last week i worked on submiting sponsors quarterly proposal, otf full proposal and started the release report.. also one on how we did on april (will put them in the wiki and email the list)
17:14:36 <isis> i need to fix the testvector code to use the constant-time `a` generation that we came up with, then add in the testvectors
17:14:37 <Yawning> nickm: "not analized nearly enough, and slow as mollasses"
17:14:59 <Yawning> (optimized shared secret generation on a Broadwel i7 is ~25 ms)
17:15:28 <teor> Oh yeah, I am still collecting responses from relay operators who want to run fallbacks (they will get their chance in 0.2.9)
17:15:33 <isis> we got the CT sampling overhead down to 4000 cycles (added) and then yawning's 5q trick removed about 20000 cycles, so it's faster and constant-time now :)
17:15:35 <isabela> (btw is great to see people adding their names on fields that is important and helpful, please do more :)
17:15:51 <Yawning> isis: glad I was useful >.>
17:16:11 <isis> Yawning: <3 many thanks, yes
17:17:25 <nickm> anybody else with an update? :)
17:17:42 <nickm> Yawning: yeah, it didn't look fast.  It was cute though.
17:18:03 <mikeperry> I am here, but I am stalled on core-tor stuff. currently doing sandboxing background research for Tor Browser
17:18:48 <Yawning> nickm: my view is "might be usable in 5 years"
17:18:59 <isabela> nickm: i will ping you to review sponsor u april report and update the spreadsheet
17:19:13 <nickm> ok
17:19:20 <nickm> isabela: time needed on that?
17:19:22 <mikeperry> (does anyone know about windows sandboxes here, btw? I am reading things that make it seem like a deep rabbit hole)
17:19:36 <Yawning> I used sandboxie
17:19:39 <isabela> we should get this out by wed max
17:19:44 <Yawning> but I do not know how it works
17:19:44 <isabela> nickm: ^
17:19:53 <nickm> mikeperry: AFAIK it is all based on installing supervisor processes as debuggers for others, but that could have changed.
17:20:08 <Yawning> and it could have just been presenting a fancy dialog and not doing anything for all I know
17:20:09 <nickm> #action nickm figure out how sandboxie works, talk with mikeperry about sandboxes
17:20:18 <nickm> #action nickm answer isabela's questions about sponsor U april stuff
17:21:15 <Yawning> (fuck it, just run tails in virtualbox if you really want "sandboxing on windows"?)
17:21:37 <isis> O_o
17:23:17 <nickm> any more updates?
17:23:28 <nickm> if not, any discussion items?
17:24:11 <isis> are there any changes to my proposal i should make?  does it get a number?
17:24:29 <nickm> Does everybody know about the https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam#Usefulticketqueries
17:24:46 <isis> do we have funding for moving forward on implementing the other proposals which the PQ one depends upon?
17:24:57 <nickm> #action somebody give isis's proposal a number
17:25:36 <nickm> I think we have a proposal out which, if granted, should give us funding for a pretty broad crypto revamp
17:25:45 <nickm> and in general, "do x" includes "do the prerequisites for x"
17:26:19 <isis> because implementing prop#249 and prop#264 could start happening soon/now, then it's there when we finally agree on a handshake
17:26:24 <isabela> nickm: those queries are great, thanks for creating them
17:26:36 <isis> nickm: okay, cool
17:26:45 <nickm> we also have an open grant proposal that covers 264 specifically
17:26:51 <nickm> in both cases we'll see if it's approved
17:26:58 <isabela> that is why it's helpful for folks to make sure they are filling up fields on tickets
17:30:19 <nickm> any other topics for this week?
17:31:09 <nickm> isabela: are there any fields in particualr that your'e thinking about?
17:31:13 <nickm> *sp
17:32:18 <isabela> reviewer
17:32:19 <isabela> owner
17:33:24 <isabela> points
17:33:33 <nickm> ok. I'll encourage folks around here to do that.
17:33:36 <isabela> nickm: can we update the query to have points column
17:33:43 <nickm> isabela: feel free!
17:33:43 <isabela> i can do it
17:33:44 <nickm> :)
17:33:47 <isabela> cool!
17:33:48 <isabela> tx
17:34:07 <nickm> I'll encourage folks here in this room to set those
17:34:09 <isis> isabela: i didn't understand why "3 points" is double the work of "2 points"
17:34:18 <nickm> #action everybody set owner, reviewer, points
17:34:23 <nickm> isis: we're thinking of 1 point == 1 day
17:34:28 <nickm> if that's okay with everybody
17:34:29 <Yawning> isis: it's a log scale?
17:34:34 <nickm> to make addition work
17:34:42 <isis> ah, okay, i see further in the thread
17:34:52 <Yawning> >.>
17:34:58 <isis> i think i like 1 day == 1 point better
17:35:31 <isis> otherwise having one person do 6 points and the other do 8, but really they worked the same amount seems weird
17:36:13 <isis> (totally biased since usually the tickets i take are "3 points")
17:36:20 <isabela> is because we are calculating the time period
17:36:34 <isabela> the goal here is to see if the person is overwhelmed
17:36:36 <armadev> isabela: so as an example, for #18809, what should be the reviewer, owner, and points fields?
17:37:17 <armadev> reviewer is currently teor, who hasn't reviewed it yet. but it got reviews from dgoulet and andrae.
17:37:19 <armadev> (andrea, too)
17:37:35 <armadev> it's owned by andrea, who set herself as that because i guess she was going to review it.
17:37:44 <isabela> owner who worked on the code, reviewer who reviewed it, points we have have a scale at release guidelines
17:37:55 <isabela> i see teor as reviewer
17:39:02 <isabela> the goal of all these things is to help visualize what is going on so we can balance work and avoid having people overwhelmed.. if you are doing something you shoud put your name in the field
17:39:23 <isabela> nickm's queries shows how the load is distributed
17:39:44 <isabela> ideally the team should review this distribution during meetings like this
17:39:51 <isabela> to help each other and create a good balance :)
17:40:10 <nickm> the may-by-owner and may-by-reviewer ones are a bit unbalanced
17:41:09 <nickm> reviewer isn't too bad actually
17:41:17 <nickm> though I hope that the reviewer===none ones get picked up
17:42:52 <nickm> or I could review my own darned tickets, but wow that's not a good practice. :)
17:42:56 <isabela> does ppl have any questions or feedback on this?
17:43:54 <nickm> athena, mikeperry: I'm calling #18365 merge_ready based on people's comments.
17:44:06 <nickm> please say on the ticket if you disagree
17:45:09 <nickm> I think we can endmeeting if there's no other discussion?
17:45:15 <teor> isabela: I review some tickets as they come in, but sometimes forget to put myself as reviewer
17:45:16 <nickm> people in this room are all distracted :)
17:46:16 <isabela> teor: is a habit thing that is why we check in on the meeting and remind folks
17:46:23 <isabela> nickm: ! hehe
17:46:28 <nickm> ok.
17:46:29 <nickm> #endmeeting