#saltr (irc2p) | I2P+ 2.6.5+ >> i2pplus.github.io | skank.i2p | git.skank.i2p | purokishi.i2p | vituperative.github.io/i2pchat | natter.i2p | notbob.i2p | ramble.i2p | shreddit.i2p | cake.i2p | /msg an op for voice

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 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? :)

dr|z3d wordnik.i2p/words/tulle

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