#ls2 (irc2p) | #ls2 I2P Protocol Development Discussion - Next Meeting Monday, October 9th 6:30 UTC

eyedeekay OK just a few things of note on that then. I started trying to get x25519 implemented in the go-i2p code, still a little trouble I guess seeing where crypto and protocols fits together to make something that does something instead of just writing a ton of unit tests

eyedeekay But I suppose that is why it's going to pay off to work on NTCP2 and the underlying crypto required to do it

zzz well, if you're getting to the point of talking to other routers, just trying to actually do it may be easier than unit tests.

eyedeekay It's not much yet, I'm at a stage of figuring out what the interfaces need to be still but I feel like I'm moving quicker than I did on the refactor

zzz Unit tests may be more disciplined, but may or may not be the most efficient

eyedeekay I did get a reliable offer to help from Keith who does the Go syndie implementation and works on some other Go software in our ecosystem, he's a pretty reliable contact and reached out to me by email

eyedeekay Nothing works yet but maybe I actually try to talk to another router in a very basic way this month

zzz I also have the very old NTCP2 standalone test code that may be of some use as the other end

eyedeekay Yeah you shared some of it with me already, I've got a big directory full of samples

eyedeekay Some of them were already helpful when I had to refactor all those tests

zzz problem with live net testing is doing any logging gets really noisy

zzz you could also try java router as the other side on a testnet

zzz lots of possibilities

zzz live net is good at telling you it worked, bad at telling you why it didn't

orignal maybe he should start with SSU2 instead

zzz no, I think SSU2 is about 3x harder

eyedeekay My plan is essentially testnet, I've got hardware to put it inside and keep it away from the main net and build a small model with only a few routers

eyedeekay I don't know LXC as well as I do Docker but I'll manage

eyedeekay I'm already wrist-deep in NTCP2, probably do more harm than good to change course

zzz there's just _so_ much to deal with in UDP protocols that isn't an issue with sockets

zzz anything else on 1a) go-i2p?

zzz or anything on 1b) monerocon?

eyedeekay Nothing on go-i2p

eyedeekay On MoneroCon the only thing to add is that I'm sticking up my workshop materials

eyedeekay Sort of doing it backward, I posted the interactive part first where we made a Monero I2P plugin with ShellService as a manager

eyedeekay The first and second part were less interactive and I'll post them later tonight

eyedeekay On I2PTunnel and SAMv3

eyedeekay EOT

zzz oh, speaking of SAM, some discussion while you were gone about domain sockets for SAM

zzz I don't really see the point but if you have the backlog you might want to take a look

eyedeekay Is it on this channel? I'll check major if it is

eyedeekay I can think of some cases for such a thing but I'll have to see the discussion

zzz there was a fair amount here, maybe some on ilita also. there's also an i2pd ticket

zzz orignal, you have any summary of the domain socket discussion for him?

eyedeekay Found the ticket, it's right on top of their issue tracker github.com/PurpleI2P/i2pd/issues/1751

zzz one justification was MTU, but localhost MTU is already huge. As a result of the discussion I added a section about it to the spec

orignal we decided to not do it for now

zzz the remaining justification was - maybe - added security

zzz note that it's a huge PITA on java < 16

zzz except android ofc

zzz anything else on 1b) monerocon or 1c) sam domain sockets?

orignal at least not now

eyedeekay Android domain sockets are a PITA for other reasons

orignal oh btw I have a question about SAM

zzz go ahead

orignal do we send error messages back in case of Datagram?

orignal say if session not found

orignal because polistern complained about it

zzz no idea. eyedeekay ?

eyedeekay Just a moment while I double-check...

orignal the specs say nothing

orignal so polistern says

orignal he session got disconnected

orignal but she keeps sending datagrams and doesn't know they can't go through

zzz does polistern have a proposal to improve it?

zzz I can't think of a good way to do an asynchronous error delivery

orignal her proposal is simple

eyedeekay Nothing comes across from the SAM side, all the error handling I have available happens either during the session setup or after reading a datagram

zzz control channel? error datagram?

orignal her wants me to fix crashes ))

orignal but if I check for session I don't want to be silent by send something back

orignal what if no control channel anymore?

zzz but if you crash, the control socket gets closed, so she should know that

orignal no, her contyrol socket get closed

orignal she keeps sending datagram and I crash

orignal I know why I crash

zzz lol

orignal but want to handle this istuation propely

orignal e.g it's a bug ofc

orignal but how to fix it

zzz there also is control socket ping/pong in v3.3, don't know if that would help

zzz I don't really have any ideas

zzz other than don't crash )))

zzz eyedeekay, anything else to add?

eyedeekay Not on the topic of figuring out when your datagram socket failed

zzz 2) SSU2 status

zzz orignal, your turn to go first

orignal so we send nothing back

eyedeekay Just one thing about Android however, basically domain sockets have to be used by your app or an app which a) begins with your namespace and b) is signed by the same keys of your app

orignal so I have fixed most of peer test issues

eyedeekay So if you do that with SAM it breaks all third-party apps. EOT

orignal and receive msg 7 most of the time

orignal for relay I got relay reponse and start connecting

orignal will finish it today or tommorow

orignal other side is reamining

zzz great

zzz I've been working on MTU issues

zzz and adjusting PMTU, and using up to 1500 for SSU2 addresses

zzz adding more checks to peer test as discussed last couple weeks

zzz I have concerns about whether fragmented messages are working with i2pd

zzz need to make sure all that works well

zzz eot

orignal yes, MTU is the next

orignal it seems they didn't work about wrong mtu

orignal *due

orignal and yes I will test

zzz can't ever be over the mtu. 1 byte over is an error. I'm still not sure I have it all correct

zzz still doing testing and watching the logs

orignal but you saw 1517

orignal technically datagram can be split and then merged back

orignal depending on equipment

zzz right. haven't checked recently

zzz right.

orignal but after this fix

orignal I see error about missing fragments rarely

zzz ok

orignal let's see

orignal also going to use SSU2 peer test for ipv6

orignal instead SSU

orignal you know why

zzz right, we designed it to be better

zzz it will be better if it works :)

zzz anything else on 2) ?

zzz anything else for the meeting ?

zzz 4th of july meeting ok?

eyedeekay I can make it

orignal yes

orignal I'm fine

orignal peer test seems working well now