#saltr (irc2p) | I2P+ 2.7.0-9+ >> 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

dr|z3d we could all bang on about how things are broken, but it doesn't help fix the problem. you know. yes, it's frustrating, no, it's not helpful without specifics.

not_bob I agree, which is why I try to limit how much I do it.

not_bob Anyway, hosts are, or will be in sync now. That fixed it.

dr|z3d if (not_bob is on a rant) {drop = true}

not_bob I'm done at this point :)

dr|z3d good man.

not_bob Now, where is my emotional support kitten?

zzz speaking of excessive lookups, we added this throttle a decade ago and it's still dropping huge numbers of them:

zzz 04/30 14:32:46.618 WARN [ handler 1/1] nel.OutboundMessageDistributor: Drop msg at OBEP (new conn throttle) to [Hash: 1Agi~RrrAd-r-2hKj3Ck99mvCKcBuheNWf7pSHLRmfI=] [DatabaseLookupMessage:

zzz 04/30 14:32:47.576 ^^^ 35 similar messages omitted ^^^

orignal real lookups or fakes?

zzz how would I tell the difference?

orignal I'm asking what you are printing

orignal same ident many times or different idents?

eyedeekay zzz I filed the revert MR for the schedulable blocklist deletion: git.idk.i2p/i2p-hackers/i2p.i2p/-/merge_requests/192 it pretty straightforwardly rips out the feature and nothing else, let me know if you want to eyeball it otherwise I'll just merge it

zzz different

orignal then why do you trottle?

zzz eyedeekay, I'll look real quick, why not

orignal maybe real requests

eyedeekay Thanks zzz

zzz orignal, to prevent excessive "fan out", driving the obep router to connection limits

orignal and what's the limit?

zzz it's really high, stand by...

zzz probably too high

zzz 60 new connections in 30 seconds, per-OBEP

orignal 2 connections per second. It's nothing

orignal think about p2p

dr|z3d slowly, slowly, we'll get orignal onboard with the concept of throttling. :)

orignal if I run a torrent if might be hundreds

zzz what p2p app tries to connect to 120 destinations a minute?

zzz i2psnark won't do hundreds, no. Not in 30 seconds

orignal you never know

dr|z3d I think he has a pretty good idea, orignal :)

orignal in my opinion we should trorrle if exceeds 10 requests per second

dr|z3d you're not throttling anything at all right now, orignal, right?

orignal no

orignal we only limit transports queues

orignal if a message was not sent for some time we drop it

zzz the vast majority I'm tossing are RI lookups

orignal if you do it at OBEP how do you know if it's lookup?

dr|z3d I've increased the tolerances for lookups to keep orignal happy. Doesn't seem to snag many routers now.

orignal what if you have them in garlic?

zzz eyedeekay, approved, although I believe in the tile you mean MR !93, not issue #93

zzz *title

eyedeekay Yeah that is what I meant

zzz the throttle doesn't care about message type; I'm just reporting that almost all of them are unencrypted RI lookups

dr|z3d I'm sure we can reach a happy middle ground somewhere orignal where you do some throttling of lookups, and we increase the tolerances, though tbh, 30/30s or 60/90s seems like plenty to me.

orignal zzz ?

zzz looks to me like a netdb refresh at startup or something

orignal what if it's real heavy traffic?

orignal if you don't care about message type

orignal say 1 Mb/sec

zzz heavy traffic is fine. huge fanout is not

zzz just pointing out this throttle has been there for 10 years, working well

orignal what if it goes to hundreds fo destinations

orignal dude, 10 years ago I2P was slow and usuable newtork

orignal nodays people watch youtube through it

zzz you're not going to do hundreds of dests out one tunnel, and we have guidelines on how many tunnels a server should have that cover this scenario

orignal so be carefull with trottling

orignal why?

zzz we are always careful

zzz philosophy #1. Protect the router from overload

orignal say I mutlicast video to throughs of subscribers at the same time

zzz see zzz.i2p guidelines on how many tunnels you need

orignal philosophy #1. you can protect your router first but you limits must be declared

orignal my point is if you trottle something at some level

orignal this level must be published in your properties

zzz declaration is on zzz.i2p post from years ago. thottles are implementation-dependent and subject to change at any time

orignal ofc

zzz to address ongoing and evolving threats

orignal that's why it's improrant to know

zzz are you doing 100s of RI lookups per minute out a tunnel?

orignal people keep complaining that i2p can't handle traffic of few megs per sec

zzz this is not about bandwidth

orignal <zzz> the throttle doesn't care about message type

orignal that's your statemnt

orignal meaning it's applied to garlic messages

dr|z3d he's addressed the issue to some degree, zzz. he was queuing lookups and pumping them out every 15s before.

orignal dr|z3d that

zzz that's true, but no application is sending to hundreds of different places per minute

orignal 's fixed

orignal I told you an example

dr|z3d that's what I said, orignal :)

orignal video streaming service to many subscribers

orignal multicast-like

dr|z3d that's a good start.

dr|z3d if you can ensure you're not sending out > 30/30s, all the better, or 60/90s, better still.

orignal so it would be nice to publish some property

orignal real example people run online game server

zzz zzz.i2p/topics/1584

orignal minecraft or something like this

zzz One tunnel for every 20-50 clients connected in a one-minute period.

orignal bunch of players are connected to single game server

orignal and server sends updated to all of them

orignal weko can tell better about thier setup

zzz if following the recommendations it's fine

zzz I can't allow a single tunnel to drive me to connection limits

zzz it's a very basic protection, that's why it's been there 10 years

orignal then publish a proprety

zzz I'm just wondering why the huge number of lookups

orignal and will solbe the problem

dr|z3d that's probably another alien concept for i2pd. orignal, you enforce connection limits per transport?

orignal no I don't

dr|z3d huge number of lookups when single routers are hosting 400K transit tunnels.

zzz you put a property in the build request saying how many conns or lookups you want per minute, I'll accept or reject

orignal zzz, wait

orignal you receive huge number of lookup at OBEP?

orignal put property telling your trttling limits

orignal that's all

orignal if I pick RI for OBEP and find this limit too low for my needs I bypass it

zzz ok, without a property, the default is 60 in 30 seconds

orignal it's for Java but not for i2pd

zzz so are you sending out hundreds of RI lookups per minute? or not? these are not LS lookups

orignal and dr|z3d might have different limits

orignal trough a tunnel or directly?

zzz tunnel

orignal ofc no

orignal and if you seem at OBEP as llokup

orignal it means they are unencrypted

orignal that's my question. how come?

orignal if I send a lookup though a tunnel I encrypt it for FF

orignal outbound->SendTunnelDataMsgTo (nextFloodfill->GetIdentHash (), 0,

orignal i2p::garlic::WrapECIESX25519MessageForRouter (msg, nextFloodfill->GetIdentity ()->GetEncryptionPublicKey ()));

orignal that's my code to send RI lookup through a tunnel

dr|z3d seeing those more often than we should -> Dropping duplicate DatabaseLookupMessage (RI lookup)

orignal zzz question

orignal when we receive lookup and return DSRM

orignal do we include any 3 closest routers or only close than us?

orignal docs don't answer this question

orignal or even 255

orignal well not 255 but 32 to fit 1K