~dr|z3d
@RN
@RN_
@StormyCloud
@T3s|4
@T3s|4_
@eyedeekay
@orignal
@postman
@zzz
%Liorar
+FreefallHeavens
+Xeha
+acetone
+bak83
+cancername
+cumlord
+hk
+profetikla
+uop23ip
+weko
An0nm0n
Arch
Danny
DeltaOreo
Irc2PGuest21357
Irc2PGuest21881
Irc2PGuest43426
Leopold
Meow
Nausicaa
Onn4l7h
Onn4|7h
Over1
anon2
anu3
boonst
mareki2pb
not_bob_afk
plap
poriori_
shiver_
simprelay
solidx66
thetia
tr
u5657
orignal
zzz, how do you detect if tunnel or remote lease is dead in I2CP?
zzz
orignal, I2CP doesn't know about individual tunnels/leases (except when signing its own leaseset), that's the router's problem
orignal
yes, that's what I'm asking
orignal
how do you handle it at router
orignal
I think you remeber path to a destination
zzz
how does the router do routing and maintain tunnels? yes, we cache the path if it works, we pick a new path if it failed
orignal
but how do you know if i fails
orignal
that's my question
orignal
your client protocols are on client side
zzz
right. router does router stuff. client does client stuff. I2CP is always in-between. Clients do not have access to any router API or data structures
orignal
so how do you recognize failures on router's side?
orignal
requst Acks for ratchets?
zzz
that's one way, yes
zzz
also LS lookup failures, 1st hop connect failures, ...
orignal
how is it related to lookups?
orignal
you need to check if OB/remote Lease pair is functional
orignal
while lookup goes to a FF
zzz
right, so later if LS expires, and it can't get a new one, it's a failure
orignal
how often do you request Acks?
orignal
only if you are a server
orignal
for client I guess you check if your LS got confirmed but other side
orignal
the problem is that it's too rare
zzz
give me a few minutes to research, stand by...
orignal
thanks
orignal
thanks
orignal
basically not less then 60 sec
zzz
not more than 60 sec
zzz
how is the snark testing going?
orignal
I see 200 Kbs with 24 peers
orignal
it's not impressive
orignal
must be improved
orignal
that's what I'm trying to do
orignal
my I2CP implementation is too old
zzz
great, happy you're testing
orignal
but it's too slow vs. i2p tunnels
zzz
interesting
orignal
because for example in streams with switch between tunnels quickly
orignal
if we see servere delay or malfunctioning
zzz
it's hard to do comparisons because everything is different
zzz
bittorrent vs. http; 24 conns vs. 1; sone datagrams vs. none; our streaming impl. vs. yours; i2cp vs. not
orignal
if I have 24 peers and 5 tunnels in each direction
orignal
I expect like 100 Kbs per each
orignal
my question is why each peer is so slow
zzz
maybe?
orignal
with i2cp
orignal
what's a resaon we peer doesn't send like 100 kbs
Titlacahuan
I got over 200kb/s from a single peer over I2CP with i2pd yesterday in MuWire
orignal
so for 24 peers my expectation is few Mbs
zzz
you'd have to compare to a SAM bittorrent client like XD or qbittorrent
orignal
I'm not saying it's my side to blame maybe other side send stoo slow
orignal
that's why it need to be investigated
orignal
also I saw oly few Kbs uload speed
orignal
that's the main suspect
orignal
why snark doesn't pump out faster
orignal
few kbs even for 1 peer is too slow
zzz
sure. muwire probably isn't uploading to 23 other clients at once, but bittorrent is, so it has overall bandwidth limits you don't see
zzz
bittorrent is fastest if you pick an older torrent, at least a few days old, with lots of seeds and very few leeches, so the seeds are not bandwidth limited
orignal
my bandwidth is 1.5 Gbs
zzz
I'm talking the seeder bandwidth limits, not yours
orignal
it doesn't explian low upload speed
orignal
seems there are plenty of bugs
zzz
sure
zzz
you can also add ?p=2 in snark to see streaming RTT, RTO, windows, etc.
cumlord
if helpful i can open up 3 more snarks on a specific torrent for testing
zzz
the usual reason uploads are slow is because of congestion at the other end; you have plenty of upload bandwidth, but the other guy leeching is at his download bandwidth limit and/or losing packets in his tunnels
orignal
but where does this cognesion cme from?
orignal
for that side or on the way and from I2CP ?
zzz
probably in his inbound tunnels
zzz
try running XD or qbittorrent thru SAM on the same torrent at the same time, then you'll have a good side-by-side comparison
dr|z3d
make sure you've configured up/down speeds in the configs, orignal, in case that wasn't obvious.
dr|z3d
also you may want to max out the in/out tunnels.
orignal
I did
orignal
5 tulles with 3 hops
dr|z3d
5 tulles, eh? you running a brothel there? :)
orignal
tunnels
orignal
sorry typo
dr|z3d
:)
T3s|4
lols dr|z3d - I met her at a brothel, but my wife donned a beautiful tulle at our wedding :D
dr|z3d
:)
dr|z3d
maybe orignal has his own collection of tulles he'd rather not us know about :)
dr|z3d
(and matching rainbow antlers)
orignal
no, I'm doing too many things at the time
orignal
howeever keep in mind that torrenting is on of most improtant part of I2P
orignal
and should work fast
dr|z3d
indeed, indeed.
orignal
to show fuck copyrighsters
dr|z3d
to facilitate swift acquisition of linux isos you mean.
orignal
they download torrent but can't find peer's IP
orignal
ofc not
orignal
I'm a pirate
orignal
and I tried my test on "house of dragons" rather than a linux iso ))
dr|z3d
speaking of side to side testing, you might want to also test I2P+ snark standalone.
cumlord
those aught to be well seeded by now :)
orignal
it has much less peers
orignal
than house of dragons
dr|z3d
I don't mean test downloading the torrent, I mean download the torrent and run I2P+ snark.
dr|z3d
side-by-side client test.
dr|z3d
as a download test, not really that great given the size.
dr|z3d
that said, I've seen > 1MB/s on the + torrent downloads on occasion.
T3s|4
orignal: ^same here, on +
T3s|4
16 snark tunnels In/Out, 2 hops both directions; with these options: inbound.lengthVariance=1 outbound.lengthVariance=1
not_bob
I tend to get decent speeds with i2psnark+
Snowflakes
I have diagnose but I not have registered with a narcologist. I need to get up, will drink today
Snowflakes
wine
Snowflakes
because have a trouble with heart
not_bob
But, it really depends on the swarm. I've gotten faster than 500k/s enough times.
Snowflakes
another alcocol will kill me
cumlord
seem to get around 1.5MB on upload with i2psnark+, might have some time to write a script today to load up more snark instances when the average are >1MB or something
Snowflakes
dr|z3d, so them can take me to army. I will think anything
Snowflakes
10 semptember I will use drugs for comission in a bad way
Snowflakes
maybe memantine (pharma)
Snowflakes
then will get ~60 days to mental hospital
Snowflakes
I have prescription for memantine
T3s|4
orignal: if you want speed test a bunch of large file size torrents with at least 10+ seeders, have a look here: tracker2.postman.i2p/index.php?view=Main&category=1&lastactive=0&orderby=5
T3s|4
*to speed test
dr|z3d
ok, opening salvo in the translation of console and webapps into hindi available in the latest dev build. full translation of console/i2psnark, more to follow.
orignal
zzz can we send "sucess" message status without "accepted" to save one message?
orignal
in I2CP I mean
zzz
Prior to release 0.9.14, a session with i2cp.messageReliability=none could not be overridden on a per-message basis. As of release 0.9.14, in a session with i2cp.messageReliability=none, the client may request delivery of a MessageStatusMessage with the delivery success or failure by setting the nonce to a nonzero value. The router will not send the "accepted" MessageStatusMessage but it will later send the client a
zzz
MessageStatusMessage with the same nonce, and a success or failure value.
orignal
I'm asking if recieve nonce
orignal
do we have an obligation to send "accepted"
zzz
no. don't send accepted. See last sentence in above paste
orignal
never?
orignal
can I remove this code?
zzz
for i2cp.messageReliability=BestEffort you're supposed to send accepted, but that's not the default
orignal
if (params[I2CP_PARAM_MESSAGE_RELIABILITY] == "none") m_IsSendAccepted = false;
orignal
so I send accepted in not "none"
orignal
*if
zzz
correct
zzz
and a clarification from yesterday about the ratchet ack request after 60 sec... that's attached to a message that's sent. We don't have a timer and don't send a 'bare' ack request.
orignal
yes I understand you have a message to send and check if you should attach Ack reuquest or not
orignal
but what do you do on other side if you receive Ack requets but you trafic is unidirectional
orignal
because you have risk to not recive Ack back because nothing to send
orignal
also why do you always send nonce =0 ?
zzz
on the other side, we set a timer, if there's no traffic before the timer goes off we send the ack
orignal
on which level?
zzz
ratchet
orignal
need to implement it too
zzz
nonce = 0 is normal for message reliability "none". I believe we set a nonce for SYN and CLOSE in streaming, but I'd have to double check that
zzz
but if you tell me it's "always" I believe you
zzz
haven't looked at this stuff in a while
orignal
no I'm talking about snark
orignal
seems all messsages come with nonce=0