~dr|z3d
@RN
@RN_
@StormyCloud
@T3s|4_
@eyedeekay
@orignal
@postman
@zzz
%Liorar
+Atticfire
+FreefallHeavens
+Xeha
+bak83_
+cumlord
+hk
+poriori
+profetikla
+scottpedia
+uop23ip
+weko
An0nm0n
Arch
Dann
DeltaOreo
Irc2PGuest29173
Irc2PGuest46214
Irc2PGuest65329
KiloDeltaRomeo
Leopold_
Meow
Nausicaa
Onn4l7h
Onn4|7h
Over
acetone_
anon4
anu3
boonst
mareki2pb
not_bob_afk
qend-irc2p
shiver_
simprelay
solidx66
thetia
tr
u5657
woodwose
orignal
to be able to send Ack datagram with the same nonce back
orignal
to calculate RTO and window size
orignal
current UDP tunnels are not usable for wireguard tunnels
orignal
or even "options" sections
dr|z3d
enjoy the latest + update, now with compressed (but not obfuscated/mangled) javascript in the console and snark.
dr|z3d
uncompressed js still available from git.
zzz
seems like that should be at the application layer?
zzz
most application protocols have nonces or seq. nums in there already
orignal
I would prefer to have a way to add service information without toucking the payload
orignal
see, we implement a generic UDP tunnels without any knoledge about an app
orignal
all it needs to do is to deliver UDP packets to other side
orignal
an app shouldn't care about badidth, congestion, etc.
orignal
*banwidth
orignal
so, if we had an options section it would helpp a lot
zzz
any udp app has to care about bandwidth/congestion
zzz
the replies may not even come back in datagram2. in bittorrent dht, the requests are repliable but the responses are raw
orignal
for example wireguard
orignal
yes there are many combination
orignal
that's why having option sections would be nice
orignal
maybe required reponse time
orignal
basically our problem is that we can't identify when one of tunnels died and swithc to another
zzz
wireguard has congestion control in it, right?
zzz
it just "feels" like the wrong layer to me
zzz
but we can definitely add an options section
zzz
just in case
zzz
if a flag is set, then there's an options mapping
orignal
it does
orignal
but it knows nothing about i2p specific problem
orignal
like dying routers
orignal
do you have this flag yet?
zzz
no
orignal
then tell me how are we supposed to solve dead tunnels issue?
orignal
with datagrams
orignal
we do it well in streaming
zzz
please explain it more
orignal
but we always have acks there
orignal
you pick you OB and remote lease
orignal
you keep sening packets through this pair
orignal
and one intemediate router can't make it anymore
orignal
how would you know it?
orignal
in streaming if we don't have ack we swicth to another
orignal
ofc the algorithm is more complicated but conceptually it's like this
zzz
with ratchet acks. It's done at the ratchets layer, so it works for both streaming and datagrams
orignal
but in case of datagrams we can keep sening packet to nowhere for a long time
zzz
depends how often you request a ratchet ack
zzz
streaming doesn't know about tunnels anyway (at least for us)
zzz
added
orignal
how often do you send ratchets acks? once in 30 seconds?
orignal
I mean ack request
orignal
to make it righht you need to know RTO
zzz
60s
orignal
if your traffic is heavy you should switch tunnels 10+ times during this 60s
orignal
I would say you need to request it every few hundreds milliseconds but it relayy depends on RTI
zzz
it's a tradeoff of recognition vs bandwidth, you can't ever make it perfect and only lose one packet before switching
zzz
we could add a I2CP flag bit so streaming could ask ratchet to request an ack
zzz
don't know what the criterion would be though
orignal
well I keep asking about this flag for many years ))
zzz
you know my answer, write a proposal ))
zzz
throw in a 'switch tunnels' flag while you're at it
dr|z3d
*crickets*
orignal
well it's your idea now ))
dr|z3d
you need flagellating, orignal
dr|z3d
error 500 on git.idk, eyedeekay
eyedeekay
Just figuring out what I can delete...
eyedeekay
it's back
dr|z3d
latest + in a good enough state to also be pushed to the release update channel.
dr|z3d
well this is fine, using all available vcores on a system to minify js,css and html files in our javadocs.
dr|z3d
s/fine/fun
dr|z3d
single script, single pass, all resources processed in parallel.
dr|z3d
anyways, work in progress with the "new" jquery search javadoc: 2dqk3vhpjfofoplabbywryxuaucm6k5z5vqalphfpacijlnbltrq.b32.i2p