orignal
zzz, would your PQ dest honor client LS with signature type 12?
zzz
all my MLDSA support is lightly-tested and haven't re-tested in a while, but it worked last time I did
zzz
you can test against stats.i2p or zzz.i2p or my PQ dest
zzz
altonen, have you tried doing testnet testing SSU2-only with an added qdisc delay and big drop % ?
zzz
thats a torture test that will find bugs quickly
orignal
I want to try with PQ sig
zzz
have fun
altonen
zzz not yet, i need to fix a couple of issues in ssu2 first but soon(tm)
zzz
ok. even before adding drops, I recommend adding a qdisc delay of 50 ms or so which is more realistic and can keep things from going too fast
zzz
after that, add a 2% or so drop % to start catching bugs throughout, not just in transports
zzz
then switch to ssu2-only for your router, and ramp up to 10-20% drop for a brutal test
altonen
thanks for the tips!
altonen
i saw your comment/discussion with diva in the i2pd testnet tool but i'm yet to add qdisc support for my testnet
zzz
fyi java routers can get a little unhappy on a testnet without at least 10ms or so of qdisc delay
altonen
it worked alright with me, it tried emissary-emissary, emissary-i2pd and emissary-i2p
altonen
i didn't go through all the logs ofc but streaming e2e test works
zzz
mostly the issue is end-to-end in streaming where the window slams up to the max, rcv streaming buffers overrun, and CHOKEs get sent, making it all herky jerky
zzz
if you're not doing end-to-end tests it should mostly be fine w/o
altonen
i'll try pushing it to the max at some point and see how it goes & add latency for java
zzz
the qdisc support should be built-in, I just wrote two oneliner shell scripts to add/remove delay/drop because the qdisc command is impossible to remember
altonen
i wrote mine in typescript and docker so that's an extra hurdle but should be doable
zzz
for end-to-end any drop over a few % and end-to-end is hopeless because it's exponential ** 13 hops round trip, but it's fine for transport testing
altonen
yeah