Overview and Logs for the tini2p Dev Meeting Held on 2019-08-22

Geplaatst door: el00ruobuob / oneiric

Logs

<tini2p_gitlab> 0: Greetings
<tini2p_gitlab> hi
<tini2p_gitlab> 1: What's been done
<tini2p_gitlab> A lot
<tini2p_gitlab> I've more-or-less finished the tunnel message processing classes (Hop, InboundEndpoint, InboundGateway, OutboundEndpoint, OutboundGateway)
<tini2p_gitlab> still making changes as I impl tunnel management classes (InboundTunnel, OutboundTunnel, TransitTunnel, Pool, PoolManager), but the changes are fairly minor
<tini2p_gitlab> had a few big, time-consuming refactors for the processing classes, but most of those are finished now (hopefully)
<tini2p_gitlab> I am almost done with the TransitTunnel class (used for participating in Tunnels created by remote routers)
<DavidBurkett> :wave:
<tini2p_gitlab> hi @DavidBurkett, how are you?
<DavidBurkett> Great! And yourself?
<tini2p_gitlab> doing well, thanks :)
<tini2p_gitlab> welcome
<tini2p_gitlab> to the meeting
<DavidBurkett> Happy to be here. Will observe and try not to interrupt :)
<tini2p_gitlab> right on, feel free to ask or comment at will
<tini2p_gitlab> happy to have you, as always
<tini2p_gitlab> so Tunnels are almost done, and have been what I've spent all of my time on over the last two weeks
<tini2p_gitlab> 2: What's next
<tini2p_gitlab> so, the last few I2P LS2 meetings have been a bit hairy
<tini2p_gitlab> we've been discussing, mainly, changes to 144 (ECIES-X25519 end-to-end sessions)
<tini2p_gitlab> admittedly, I could have handled things much better than I did
<tini2p_gitlab> thankfully, things seem to be moving forward again, and we've concluded that reuse of ephemeral keys is a bad idea
<tini2p_gitlab> it will require some changes to my ECIES-X25519 impl, which I will do after finishing tunnels
<tini2p_gitlab> whether to follow the drop-until-secure or send-NewSession-until-secure is still to be determined
<tini2p_gitlab> drop-until-secure is what I prefer, and means routers with drop end-to-end messages, until a secure channel to the Destination is established
<tini2p_gitlab> only requires one round-trip, and support sending one 0-RTT payload in the initial NewSession and NewSessionReply messages
<tini2p_gitlab> zzz is of the opinion that drop-until-secure will not work for performance and client-related reasons
<tini2p_gitlab> I haven't begun to implement client stuff, so maybe zzz is right
<tini2p_gitlab> NewSession-until-secure basically sends NewSession messages to the Destination until the first NewSessionReply is received
<DavidBurkett> Is that discussion in the PR for 144 or something?
<DavidBurkett> Or was that in IRC?
<tini2p_gitlab> was on Irc2P, the meeting logs are here: tini2p/meta#29
<DavidBurkett> Thanks :)
<tini2p_gitlab> and here: tini2p/meta#28
<tini2p_gitlab> guess it might be a bit much to rehash the discussion here, just trying to summarize
<DavidBurkett> I'll read through and understand
<DavidBurkett> YOu have good TL;DR's
<tini2p_gitlab> NewSession-until-secure also has its problems, but should be workable
<tini2p_gitlab> thanks :)
<tini2p_gitlab> anyway, I'll be implementing the drop-until-secure until we come to consensus on what actually becomes spec
<tini2p_gitlab> it's simpler, more secure, and more efficient
<tini2p_gitlab> though latency/performance may be a killer for it
<tini2p_gitlab> after the changes for ECIES-X25519, I'll be gluing everything together with a RouterContext class
<tini2p_gitlab> with RouterContext, I'll be able to have the NTCP2 transport talk to the Tunnels and NetDB classes
<tini2p_gitlab> I'll create some basic unit tests, if they make sense. the majority of the tests for RouterContext will be net-tests (integration/functional tests)
<tini2p_gitlab> with RouterContext finished, tini2p library users will be able to build their own router
<tini2p_gitlab> that will likely be alpha-release time
<tini2p_gitlab> I'm shooting for a one-week impl time on RouterContext, but may run into bugs/refactor work that pushes that timescale back a week or two
<tini2p_gitlab> it's really just gluing stuff together though, so I'm hoping it goes quickly
<tini2p_gitlab> I also need to make a small update to NTCP2 for the testnet-separation update to specs from 147
<tini2p_gitlab> basically it's a netID check that ensures no cross-talk between mainnet and testnet routers
<tini2p_gitlab> it's important for I2P network health, so I am happy to impl
<DavidBurkett> That's awesome! I'll watch for updates on that RouterContext. I'll try setting one up once you're finished
<tini2p_gitlab> for sure :)
<tini2p_gitlab> will likely be adding a docker setup, so you can run a full testnet entirely locally
<DavidBurkett> Nice!
<tini2p_gitlab> definitely! also allows spinning up 10+ routers on different /16 pretty easily, so it may be a good way to do integration tests with I2P Java and i2pd once I add ElGamal tunnel building
<tini2p_gitlab> ElGamal tunnel building is post-alpha though
<DavidBurkett> :)
<tini2p_gitlab> so that's about it for stuff until alpha release
<tini2p_gitlab> 3: Questions/Comments
<DavidBurkett> Nothing from me. :thumbsup:
<tini2p_gitlab> for sure
<tini2p_gitlab> 4: Next meeting
<tini2p_gitlab> staying on normal schedule, two weeks from today: 2019-09-05 18:00 UTC
<tini2p_gitlab> that's all, thanks all for the meeting
<@tini2p_gitlab> gaffer banger


Post tags : Dev Diaries, Kovri I2P Router, Cryptography