13:30:02 <nickm> #startmeeting 13:30:02 <MeetBot> Meeting started Wed Aug 12 13:30:02 2015 UTC. The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:30:02 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic. 13:30:20 <nickm> good morning, friends! 13:30:36 <nickm> Let's start with quick quick status reports and move on to discussion, as is our want. 13:30:38 <teor> good <insert-timezone-here> 13:30:40 <nickm> *wont 13:31:25 <nickm> I've been working on reviewing stuff, getting all of s7r's ed25519 usability issues solved, etc 13:31:54 <nickm> and I've also been hacking away at the Blob (for blob == "our largest strongly connected component in our call graph") 13:32:37 <nickm> I hope I can get a lot of writing done, but code is so interesting! 13:32:46 <nickm> Also we need to talk about freeze for 0.2.7 13:32:55 <nickm> who's next? 13:33:07 <Yawning> debugging 13:33:22 <nickm> oh? 13:33:34 <Yawning> stupid ed25519 batch verifyciation code sigsegvs 13:33:39 <nickm> ah 13:33:48 <Yawning> uh, apart from that, did pt paperwork 13:33:57 <Yawning> now back to my regularly schedule tor fun 13:35:38 <teor> I reviewed the SR proposal and some of nickm's patches. 13:36:39 <teor> I started work on the fallback directories again, and should be able to have the whitelist code ready really soon 13:37:24 <teor> The challenge is then to get the opt-in organised and work out how many fallbacks we need 13:37:25 <teor> and the max percentage of clients we want any one fallback to see 13:37:32 <teor> #15775 13:37:39 <teor> done 13:37:52 <athena> finishing off/checkpointing a bunch of dirauth dos filter (#4581) glue and connection-counting hash table i've been on for the last few weeks before leaving for cccamp later tonight - will be testing/debugging after i'm back 13:38:25 <nickm> sounds good; is it ready for initial review? 13:40:23 <nickm> athena: ^ 13:40:33 <athena> close; still some bits missing but i might get a chunk of them done before i leave later 13:40:42 <nickm> also, anybody else with an update, or shall we move on to discussion and stuff/ 13:40:43 <nickm> ? 13:41:38 <nickm> ok, discussion! 13:41:50 <isabela> cool 13:42:32 <nickm> Discussion topics include: deliverables to do for october-ish; how to handle the 0.2.7.x freeze; and... not sure what else 13:43:03 <isabela> ok 13:43:56 <isabela> I guess the first topic is more for athena and Yawning 13:43:59 <nickm> if no other topics, let's talk deliverables 13:44:02 <isabela> since we dont have all the people here 13:44:21 <isabela> have you guys seem -> https://docs.google.com/spreadsheets/d/1dTva10mu-FcX8KrxRjgkFvHSyNy7aBpD9xehNuUeZ-4/edit#gid=0 13:44:26 <nickm> yeah. I'd love to know if others think they can pick up any of the stuff on that list with a super-high chance of getting something done on time. 13:45:38 <isabela> I ended up without time to add the tickets there but is on my list 13:45:44 <nickm> (One undesirable outcome is picking it up and then kicking it back to me at the last minute. :) ) 13:45:46 <athena> as i indicated earlier, i can definitely take row 6 but i believe it has a dependency on 5 and i'm not current enough on plans for path/guard selection changes to write that spec 13:46:20 <Yawning> I have no idea about the guard stuff 13:46:56 <athena> i think asn may be the only person who really does get that and we're short an asn this week 13:47:05 <isabela> yes 13:47:16 <isabela> athena: i will putyou on 'who can do' for row 6 13:47:22 <isabela> for now 13:47:29 <isabela> Yawning: maybe some of the documentation part? 13:47:30 <nickm> right, but let's make sure we return to the asn part of it rsn 13:47:31 <isabela> or testing 13:47:36 <Yawning> ??? 13:47:47 <isabela> Yawning: just listing other tasks not related to guards 13:47:51 <isabela> that we need help 13:48:11 <Yawning> the deadline on it would pass before I know enough about the guard thing 13:48:17 <isabela> nickm: yes, we should do this again next week / i will follow up via email and cc folks this time 13:48:30 <Yawning> I have one thing in my pipeline for more DoS resistances for HSes 13:48:45 <Yawning> which may be 13 13:48:53 <Yawning> but, it's not a propsal, just a code improvement 13:49:17 <nickm> Yawning: turning it into a proposal -- would that take more than an hour? :) 13:49:30 <Yawning> it's not really a spec change type thing 13:49:38 <Yawning> just "yeah, we drop INTRO2s if the user asks" 13:50:17 <asn> asn is currently going to CCC camp 13:50:26 <Yawning> I could go more elaborate and be like 'The HS can use the super special HSOrPort to query the app if it should accept or reject a incoming rdv attempt' 13:50:41 <Yawning> which would be proposal worthy but is a few orders of magnitude more complicated than what I ahd in mind 13:51:02 <nickm> hm. 13:51:40 <nickm> well, we can always think of another anti-DoS thing and write a proposal for it 13:51:52 <nickm> maybe something to handle socket exhaustion on exits or something. 13:52:24 <Yawning> wtffff why is this bullshit segfaulting in crypto_rand 13:52:27 <asn> it is still in my August TODO list to write the guard algorithm 13:52:28 <Yawning> *punches wall* 13:52:57 <asn> internet in this train is bad. see you later. please send mail if you want careful reply! 13:52:58 <nickm> Yawning: paste a stack trace? 13:53:24 <nickm> asn: ok. soon we'll be blocking on it, so getting it together soon will really help 13:53:31 <nickm> we're not blocking on it quite yet though 13:53:41 <Yawning> nickm: #0 0x0000555555b3c858 in crypto_rand () 13:53:43 <Yawning> #1 0x000055555582779a in ed25519_randombytes_unsafe_donna (len=64, 13:53:45 <Yawning> p=0x7fffffff6ca0) 13:53:48 <Yawning> at ../src/ext/ed25519/donna/ed25519-randombytes-custom.h:16 13:53:55 <Yawning> p is def valid 13:54:04 <nickm> and has 64 bytes? 13:54:08 <Yawning> yeah 13:54:31 <Yawning> (gdb) p sizeof(batch.r) 13:54:33 <Yawning> $2 = 2064 13:54:42 <nickm> this is happening from unit tests, or what? 13:54:50 <Yawning> yah unit tests 13:55:08 <Yawning> the code you had in crypto_ed25519.c would have segfaulted later, but I fixed that >.> 13:55:40 <Yawning> err earlier too 13:56:05 <nickm> hm. let's poke more after meeting 13:56:26 <Yawning> sorry I should focus on the meeting 13:56:36 <isabela> :) 13:57:02 <nickm> It might help if I could get somebody to help with all the documentation tasks I have for sponsorU, but I'm not sure whether I .can. delegate those. 13:57:28 <nickm> like, there are corners of tor's internals that would need serious archeology for somebody else to write about. 13:57:29 <isabela> nickm: can we do a check on this list every meeting? just so we dont miss anything in this list couple months before deadlines 13:57:29 <Yawning> I can think more about the DoS stuff but my plate is suprisingly full 13:57:34 <nickm> sounds good 13:57:47 <nickm> Yawning: yow, thanks 13:57:54 <isabela> people can pick up stuff later as they have more time too 13:58:05 <isabela> we just cant ignore this list :) 13:58:10 <athena> nickm: yeah, there are definitely significant pieces of tor i don't think anyone but you has ever touched 13:58:27 <nickm> I bet that team effort on the whitepaper about classifying DoS attacks might be smart. we're all good at those. 13:58:43 <Yawning> we did fix a bunch of gnarly DoS attacks vs HSes 13:59:01 <Yawning> but we just fixed them/added mitigations and not like "wrote propsals" 13:59:35 <Yawning> (hell, even that scalar base mult stuff is a DoS mitigation thing >.>) 14:00:13 <teor> I'm better at making DoS attacks via chutney than fixing them 14:00:22 <asn> well we wrote the shared randomness proposal. that's anti-DoS right? :) 14:01:21 <asn> #prop250 14:03:30 <nickm> maybe :) 14:03:41 <nickm> should we move on to the 0.2.7 freeze thoughts? 14:03:56 <nickm> There's a bunch of stuff I'd love to get into 0.2.7, but I'm also leery of letting anything go to long. 14:03:59 <teor> It snowed here in Canberra today 14:04:12 <teor> Oops, random association 14:04:18 <Yawning> hm 14:05:28 <isabela> nickm: maybe we should focus on what has already being selected for august? 14:05:59 <nickm> makes sense. Try to do the CoreTeam201508 stuff, and then freeze at end-of-aug? 14:06:21 <fractalcat> teor: snow in canberra? maybe i should watch the news; sydney felt a bit colder than normal. 14:06:23 <isabela> https://trac.torproject.org/projects/tor/query?keywords=~TorCoreTeam201508 14:06:33 <isabela> nickm: yes 14:06:56 <isabela> we should review to confirm if it makes sense but I think is the way to start 14:08:02 <nickm> review now, review soon, or review next week? :) 14:08:47 <isabela> do we have the time to do it now? 20min left for the 1hr meeting 14:08:51 <Yawning> still not sure if we should ship ref10 that will rarely get used 14:08:59 <Yawning> since so far no one has complained 14:09:02 <Yawning> heh 14:09:05 <teor> fractalcat: I stayed inside and coded all day 14:10:33 <nickm> Yawning: I'm a bit indifferent there. 14:11:39 <nickm> so first thing to do would be to look over isabela's list and see if there's anything assigned to you which you can't do? 14:12:03 <nickm> myself, I find myself putting off #15055 for a smarter day. 14:12:15 <asn> i have both #12595 and #16255. i hope i can do both properly. 14:12:53 <asn> i've done most of #16255 already. testing is left. but it's not easy to test. 14:13:09 <asn> i've been prioritzing on #16255 so far. but maybe i should swwitch to #12595 since so many people care about it. 14:13:33 <teor> I have been trying to help out with testing #16255, but I think there's still a bit of a gap 14:13:40 <isabela> asn: sounds like a good idea 14:13:50 <asn> isabela: ack 14:14:07 <teor> Oh, I just realised that #15775 is DoS resistance, too 14:14:41 <nickm> great 14:15:13 <teor> Is there a tag for that? 14:16:18 <Yawning> hm if freeze is in august guess I should focus on the pt stuff I want in :/ 14:16:56 <teor> If I want to get fallbacks in, they'll need to be done quick-smart. But my guess is that opt-in will push them to 0.2.7.3 14:17:03 <teor> Oops. I mean 0.2.8 14:20:12 <nickm> if it's dos, use the DoS and the SponsorU tags 14:21:08 <teor> Done. Spreading the load using fallback directories makes it harder / less productive to DoS the authorities. 14:22:58 <nickm> woo 14:23:26 <nickm> any more for today, or are folks out of energy? :) 14:24:07 <isabela> i will send an email summarizing 14:24:23 <isabela> btw dgoulet is on vacation (/me believes) and/or on his way to camp 14:24:36 <nickm> sounds good 14:24:42 <nickm> and now it's time for me to say...... 14:24:45 <nickm> #endmeeting