#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

weko how many RAM java i2p use on linux? Have some ways to optimize? I need minimum consumption for testnet.

weko ping me in answer

not_bob I think the default heap is 256m.

not_bob But, zzz or dr|z3d would be able to give a better answer.

zzz weko, edit wrapper.config, look for maxmemory

weko so 50MB for example is good value for it?

dr|z3d no

dr|z3d minimum 128M, possibly 192M.

dr|z3d have a look at /configwebapps and /congfigclients weko and disable everything apart from the routerconsole and application tunnels for a super-minimal router.

dr|z3d what you have to bear in mind is that the jvm itself needs around 64M just to start.

dr|z3d also make sure that you stop the router and then start it once you've edited wrapper.config, as a restart from the router UI or i2prouter won't pick up the changes to wrapper.config

zzz should work fine with 50 at low-medium bandwidth weko

dr|z3d actually, I'm trying to remember if 64M is the right value for JVM, maybe that's off mark.

zzz no use guessing weko, just try it, if it doesn't work you will know

dr|z3d iirc, running router.jar from the command line needs 10-12MB of ram allocated, but the JVM will report ~64MB. been a while though.

dr|z3d never got an ack from you on that full text java search engine, zzz, did you take a look at it?

dr|z3d github.com/iboxdb/ftserver

dr|z3d (maybe my messages got lost in the ether)

zzz dr|z3d, no, that topic has fallen completely off my list

dr|z3d looks promising

dr|z3d much more promising that solr/lucence, at least.

dr|z3d *than

dr|z3d typing for shit today. *lucene

snex i like elastic

dr|z3d I think that's substantially heavier than the ftserver project I linked, snex. needs to be lightweight.

snex the code looks lightweight but it looks like its storing your db in a very inefficient and unsafe form

dr|z3d iboxdb.com ?

dr|z3d both ftservr and elastic use nosql.

snex ive never even heard of iboxdb

dr|z3d well don't diss it before you've had a look :)

snex i2p on rust needs more love

snex java is the new fortran

dr|z3d not by a long shot.

weko zzz, dr|z3d, not_bob thanks

weko zzz: why SSU2 session termination timeout is 5.5 minutes?

zzz I don't remember

zzz there may be some discussion in prop. 159

orignal zzz what would happen if we send HolePunch through etsablished session?

orignal if we really need to refresh hole punch?

zzz ??

zzz you're trying to keep the firewall open?

weko continue hole punch when no any packets

zzz you're trying to keep the firewall open?

orignal yes

orignal while session if alive

orignal it lasts 5.5 minutes of incativity

orignal while hole punch might end in 30 seconds

zzz spec says:

zzz Use as Ping/Pong

zzz Path Validation and Path Response blocks may be used at any time as Ping/Pong packets. Reception of a Path Validation block does not change any state at the receiver, unless received from a different IP/port.

zzz but we don't do that

zzz we just send padding-only data packet

orignal Path validation messager?

orignal yes we do it too

zzz yes

orignal but padding only data packet changes idle time

orignal we want to avoid it

zzz so only change idle time if there's data in there

weko <orignal> but padding only data packet changes idle time

weko ye and session never ending

orignal we want to send a packet that doesn't change last activity time

orignal zzz ?

orignal really?

zzz then use path validation/response

orignal then how do you handle keep alive with introducer?

orignal I send padding only data packet

zzz same

orignal and you claim you don't count it

orignal for idle

zzz looking...

orignal it means if you are an introducer you should terminate the session

orignal while you are looking

orignal Would you honor FF with G cap?

zzz /**

zzz * When did we last receive a packet from them?

zzz * Set for data, relay, and peer test, but not acks, pings, or termination

zzz */

zzz void setLastReceiveTime(long when) { _lastReceiveTime = when; }

orignal same in i2pd

orignal and we need to simething that doesn't change last receive time

orignal send

zzz so only change it for I2NP blocks

orignal then what do you do for introducers?

orignal you need change last activity time

zzz we ping our introducers

orignal how?

zzz padding-only data packet

orignal <zzz> so only change it for I2NP blocks

orignal no I2NP block no change of last receive time. right?

orignal I'm Alice and send padding only data block

orignal and your Bob closes the session soon

orignal bug

zzz actually I have a bug, I'm not updating receive time at all (((

orignal yes, that's bad

orignal I update receive time on every data packet regardless what's inside

zzz yeah I think that's probably right

zzz or maybe not

zzz cause you shouldn't update if it's just an ack

orignal then we need to decide what we send as keep-alive

orignal what's wrong with ack?

orignal ack means a reponse to your message'

orignal it can't be unsolicited

zzz we keep track of both last-sent and last-received, idle time calculation is based on the max of those

orignal yes, but ack doesn't make much difference

zzz but we use a longer idle timeout if we offered to introduce. why not do that?

orignal anyway it doesn't solve the problem with introducers

zzz then use path challenge/response

orignal huh? you keep session active for an hour if iTag?

orignal even if no acivity there?

orignal no, I'm asking about keep alive

zzz we have a 20 minute idle timeout if we offered to introduce, yes

orignal and terminate after 20 mintes?

orignal if no activity if with keep-alive

zzz correct

orignal doesn't sound right

orignal how can I make a session last for 1 hour then?

zzz if you're actually using the introducer, any relay block will count as activity

zzz weko, there is no 5.5 minute timeout recommendation in the spec, and that's not what we do

orignal it might come after 20 minutes

orignal that's the problem

orignal then what's your temination timeout now?

zzz does the spec say it's guaranteed for 1 hour?

orignal I copied from SSU1

orignal it should say this

zzz I don't think it was specified in SSU1 either

zzz are you trying to keep your firewall mapping open, or keep the SSU2 connection open?

orignal two different tasks

orignal in case of introducer I'm trying to maintain the seession

zzz we use padding-only packets to keep the firewall open

orignal hmm, I assume we keep session alive this way

zzz there is no way to force the session to stay open. there's no minimum idle time in the spec

orignal 1 hour was somewhere

orignal it's not my invention

zzz I could set my idle time to 10 seconds in theory. But if I sent you an itag, I set it to 20 minutes

orignal we do it deifefrently however

orignal if no activity still same 5.5 minutes

zzz and after 1 hour do you stop pinging?

orignal yes, we don't use it as introducer anymore

orignal ofc if some data goes through it will stay alive

orignal if not, termination after 5.5 minutes as usual

zzz our minimum timeout right now is 165 seconds

zzz anyway, we count path challenge/response as activity, and that's what the spec says to use for ping/pong, so use that if you want

orignal fine

zzz re: fG routers, we don't care about congestion caps for floodfills

orignal great

orignal we have this idea

orignal becuase floodfills seem malfunctioning due to excessive transit

zzz maybe I'll change to path challenge also for pings. I'll think about it

zzz malfunctioning how?

orignal too much traffic

orignal people think becase too much transit

zzz and they publish G?

orignal they want to set "no transit"

orignal but it will publish G

orignal my FF is almost always E or D now

zzz better to fix the malfunction ))

orignal at 80 Mbs cap

orignal no it's not fixable mulfunction

orignal some people want to run a floodfill but without excessive transit

zzz then set the limit to 1000 or 100, don't set it to 0

orignal they have a problem with network

orignal O is minimal for FF in i2pd

orignal and some people don't want even O

orignal e.g transit with 256 cap

zzz I mean zero, not capital O

orignal I know what you mean

zzz set transit limit to 1000, not zero

orignal if you set to 100 it will publish L

orignal even 256 is too much for them

zzz why? bandwidth != tunnel count

orignal are you talking about bandwidth or count?

dr|z3d so you want some sort of load balancer for floodfills.. either turn floodfill off if G, or reduce the max # of transit tunnels.

zzz count

orignal yes it's another option

orignal but somethimes a single tunnel produces 10 mbs

orignal that's the problem

zzz don't accept tunnel if you don't have available bandwidth

orignal simply speaking count means almost nothing now

orignal right that's what happene

orignal they always at O cap

zzz do you calculate average bandwidth per tunnel?

orignal no

orignal no demand for it

orignal I calculate overall bandwidth

orignal Transit Tunnels: 42196

orignal voila

orignal how do you like it?

zzz that's how you know if you can accept one more tunnel. If (current bw + avgpertunnel) > max part. bw, then reject tunnel

orignal I bumped number of transiot tunnes limit to 50K this morning

orignal and what you do if an existing tunnel start trnsmiiting too much data?

orignal when you accepted it you were good

zzz we start dropping

zzz and rejecting

orignal that's what we do now

dr|z3d last time we had a discussion about transit tunnel count, 10K was considered high, 13K was considered wrong.

zzz you can't "fix" congestion by getting rid of limits

orignal let me set the limit to 100K

orignal looks like the attack if about to stop all router accept new tunnels

dr|z3d yeah, that looks about right, orignal. transit tunnel saturation.

orignal and since your drop is based to TBM ratio

orignal rate is very low now

orignal because you can't differentite shit and legit TBM

dr|z3d ~54% here on one router. not great, not calamitous either.

dr|z3d (tunnel build success)

dr|z3d but you were talking about something else, sorry.

orignal same of firewalled one

orignal but very low on a high loaded floodfill

orignal because you guys like to drop my TBM

snex snarkFoot only displays if size() > 1. why not size() > 0?

snex the display logic even handles singular vs plural wording lol

weko zzz: do you use your own router for tunnel ends?

zzz of course not

weko zzz: so then there is deanon attack. do you know how enigma was broken?

weko there we can use same attack

weko if some router in b32's tunnels ending, we know what this router is not b32's owner

zzz so do it then, you don't need my permission

dr|z3d snex: off the top of my head, > 1 is because you don't get any meaningful information for 1 torrent that isn't already present in the table.

dr|z3d haven't looked at that code in a while, anyways, I'll see if > 0 makes more sense.