#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

parabo I sent men into Ukraine. They were wearing USSR flags. They were unarmed, they have been telling Ukrainians if they don't fight when Russia takes their region they will be free to stay or leave, up to them which

GalaxyNova ?

parabo Because when Putin takes a town or city, he doesn't do anything to non-combatants

parabo he basically just says okay you're russian now, you can stay or leave

parabo but if you are fighting, you will be executed

GalaxyNova parabo: you sent soldiers to ukraine?

parabo no, not soldiers

parabo just men\

GalaxyNova oh

GalaxyNova to like help with supplies?

GalaxyNova that's pretty cool

parabo not supplies

dr|z3d what planet are you residing on, parabo?

parabo to tell the ukranian soldiers if they stop fighting they won't be hurt

parabo Earth\

dr|z3d putin is a war criminal.

itsjustme yeah, I've not seen much good from putin

GalaxyNova oh so you're a russian shill

parabo GalaxyNova, no I'm trying to make them not die

parabo because obviously Putin will win

parabo Whether we like it or not

itsjustme dr|z3d: I think you need to watch RT.com to get the (real) story :P

dr|z3d civilians are getting murdered en masse.

dr|z3d haha, itsjustme

parabo dr|z3d, the pentagon released a statement that all but the bombing of Kiev, they can't confirm any of that

parabo but I am not on russia's side

itsjustme is it even a war?

parabo I'm trying to prevent death

itsjustme I heard it was a "special operation" :P

GalaxyNova Someone should make an eepsite with unbiased news on Ukraine

itsjustme SPECIAL OPERATION!

parabo there's no such thing as unbiased news

GalaxyNova there's reporting on objective facts

GalaxyNova without all that urban legend ghost of kiev stuff

parabo according to the pentagon, all but the bombing of Kiev, cannot be proven, as far as war crimes go

parabo I don't like Putin

parabo I'm not defending him

itsjustme *** is going to sit this debate out ***

parabo I'm sharing what I know and have heard

GalaxyNova hm

parabo Some conspiracy theorists have theorized that Joseph Stalin, may have been a communist

GalaxyNova haha

mesh the good news is that even after a week of uptime the i2p router doesn't report consuming much moe than 400mb of memory

mesh according to /graphs

GalaxyNova that's good

mesh according to htop of course the real memory usage is ~2.4gb

mesh this is despite being a floodfill and participating in hundreds of tunnels and having 1500+ active connections

GalaxyNova it's running i2p+ i assume?

mesh the cost of this is high cpu usage. Between using ZGC, which consumes a lot of cpu, and all the encryption work that I2P does, we see cpu usage spiking 10-30%

mesh GalaxyNova: latest i2p+ release

mesh It seems you don't need that much memory to run a I2P floodfill. Probably 4gb would be enough. But you probably do want a faster cpu and faster ram. I2P floodfill router is a low-latency, compute intensive application

GalaxyNova what does a floodfill router do differently from a normal router?

GalaxyNova does it use as many resources as it can?

mesh it participates in the floodfill dht and it ends up having lots of long-lived connections to other floodfills and transmitting a significant amount of data back and forth between them about the state of the i2p network

mesh floodfills by their very nature also end up participating in lots of tunnels

GalaxyNova oh i see

mesh man I can't believe how awful the digital ocean website has become

mesh you could probably a floodfill router fine on a basic digital ocean droplet with 4gb/2cpu/4TB for USD$24/mo

GalaxyNova I wonder if a physical i2p router could exist

GalaxyNova then you woudn't need to install anything on devices

mesh GalaxyNova: it would just be a computer running java in a shiny box

mesh GalaxyNova: but yes I like that idea too

mesh you could also probably run a floodfill router on a hetzner CPX31 (8gb,4vcpu,20tb) instance for ~USD$13/mo but better would be a CCX12 (8gb,20tb,2dvcpu) for ~USD$20/mo

dr|z3d 4GB for a floodfill? that's overkill.

GalaxyNova Is it possible to run a i2p tunnel to an arbitrary website you don't own? Like if I wanted to make a tunnel to wikipedia.org would that work?

mesh I conclude that the cost of running a floodfill is probably not that great, ~USD$15-20/mo

mesh dr|z3d: I've been running a floodfill for about a week that consumes ~2.5gb of ram. I'm thinking you want at least 4gb ram on your server

dr|z3d mesh: you're running java unconstrained, memory-wise?

mesh $JAVA_HOME/bin/java -Di2p.dir.base=$I2P_INSTALL_DIR -Di2p.dir.pid=. -Di2p.dir.tmp=./data/tmp -Di2p.dir.config=./data/config -Djava.library.path="$I2P_INS

mesh err oops

mesh $JAVA_HOME/bin/java -Di2p.dir.base=$I2P_INSTALL_DIR -Di2p.dir.pid=$I2P_WORKSPACE_DIR -Di2p.dir.tmp=$I2P_WORKSPACE_DIR/data/tmp -Di2p.dir.config=$I2P_WORKSPACE_DIR/data/config -Djava.library.path="$I2P_INSTALL_DIR/lib" -cp "$I2P_INSTALL_DIR/lib/*" -XX:+UseZGC -XX:SoftMaxHeapSize=4G -XX:+UseNUMA -XX:+UseStringDeduplication net.i2p.router.Router

mesh that's the command I use to run I2P

dr|z3d so softheap is 4GB..

dr|z3d you can run a floodfill with 384MB RAM allocated to the jvm.

mesh I don't set a max heap size. To give zgc best performance just set a softheapmaxsize. But it doesn't really matter either way the server cconsumes about 2.5gb of ram pretty consistently

dr|z3d java, however, will use whatever it can, sometimes reported by htop as much more than allocated. not entirely sure htop is accurate here.

mesh dr|z3d: there's a difference between heap size and the amount of memory the jvm consumes. If you look at htop I doubt you'll see java consuming 384mb of ram

mesh the problem is that the heap is just one of the memory pools the jvm allocates. There's also metapsace and memory mapped files

dr|z3d indeed, mesh. I'm just not entirely confident htop has it right, either. but i2p/i2p+ will run quite happily as a floodfill with 384MB allocated to the JVM on a 1GB system. whether it runs optimally is another issue, but it'll run fine enough.

term99 i have ran i2p+ on my rpi3 and my rpi4, can confirm both versions run smooth rpi3 has heavier load but still runs

mesh dr|z3d: my heap size rarely crosses 400mb though there are spikes to as high as 1gb. But heap size is only a part of the total memory usage. I think the simple size of the codebase consumes a bit

dr|z3d and bear in mind that paging out to swap isn't a major issue, either.. plenty of stuff that's held in ram can be swapped out without a major degradation in performance.

mesh dr|z3d: yeah I'd be curious to gather metrcs on such a constrained system. running a i2p floodfill with 1gb of ram would I think lead to a very busy garbage collector and bad performance

mesh dr|z3d: but yeah you could probably run a floodfill on a potato. I would say optimal settings might be something like 4gb ram and 2 dedicated cpu (or maybe 4 vcpus) on modern chips

mesh especially if you want to actually saturate your bandwidth

mesh I significantly increased the bandwidth settings from the default which might explain the 2.5gb ram usage

mesh but that's kind of the point

dr|z3d not entirely sure saturating your bandwidth is possible for the most part. there were some upstream changes relating to CoDel a while back that seemed to really throttle the router. prior to that, we were regularly seeing 10MB/s participating tunnel traffic.

mesh what is CoDel?

dr|z3d it's a queue management method. aka Controlled Delay.

dr|z3d en.wikipedia.org/wiki/CoDel

mesh very interesting

mesh it makes sense. better latency at the cost of less throughput

dr|z3d well, I've been trying to restore the previous throughput for the last 6 months or more, with limited success :|

mesh dr|z3d: that reminds me, do you know what the router does to support "Interactive" vs "Bulk" tunnels?

mesh dr|z3d: It's always going to be a tradeoff. I suspect for many people latency is more important than throughput. But then you've got the torrenters who probably want to move as much data as possible

dr|z3d I think interactive lowers the reponse time of the tunnel.

mesh dr|z3d: do you have any idea how that translates into the I2PSocket/Options api?

dr|z3d if ("interactive".equals(_profile))

dr|z3d // This was 1 which doesn't make much sense

dr|z3d // The real way to make it interactive is to make the streaming lib

dr|z3d // MessageInputStream flush faster but there's no option for that yet,

dr|z3d // Setting it to 16 instead of the default but not sure what good that is either.

dr|z3d config.setProperty("option.i2p.streaming.maxWindowSize", "16");

dr|z3d else

dr|z3d config.remove("option.i2p.streaming.maxWindowSize");

dr|z3d so it reduces the transmit window, it appears. default is 128.

mesh ah thanks, that makes sense

dr|z3d public boolean isInteractive(int tunnel) {

dr|z3d return getProperty(tunnel, "i2p.streaming.maxWindowSize", 128) == 16;

dr|z3d }

mesh it's too bad most academic research is focused on tor

dr|z3d it is what it is!

mesh nah, i2p has much more potential than tor. in fact tor should be rewritten on top of i2p

dr|z3d the more popular i2p becomes, the more likely people will want to work on it.

mesh there's no reason tor directory servers couldn't just be a special kind of i2p router

dr|z3d no harm in blue sky thinking :)

mesh dr|z3d: I mean logically packet-switched networks are better than circuit-switched networks

mesh I would expect i2p to win over tor, for the same reason the packet-switched IP protocol won

mesh and if I2P were to relax some of its constraints (tunnels only last ~8mins) you could def implement tor functionality on top of i2p

mesh that's the only real difference between i2p and tor I think. tor circuits are long lived and i2p tunnels are actually very short lived

dr|z3d actually, for the most part, circuits and tunnels are roughly the same, time-wise. maxCircuitDirtiness determines the default circuit length, and is 10m by default.

dr|z3d also remember that tor circuits are bi-directional, i2p tunnels are uni-directional.

mesh it would be interesting if i2p had support for long lived tunnels (assuming each peer agrees). like if you're sending a bunch of data and everybody agrees you could negotiate 60minute tunnel lifetimes

mesh dr|z3d: I don't think that's how it works. I think a tor circuit will live as long as data is flowing over it right? Unless this changed

mesh or as long as there's an active connection over it. So if you use irc over a tor circuit that tor circuit might last for weeks

mesh support.torproject.org/#ChangePaths

mesh Tor will reuse the same circuit for new TCP streams for 10 minutes, as long as the circuit is working fine. (If the circuit fails, Tor will switch to a new circuit immediately.)

mesh But note that a single TCP stream (e.g. a long IRC connection) will stay on the same circuit forever. We don't rotate individual streams from one circuit to the next. Otherwise, an adversary with a partial view of the network would be given many chances over time to link you to your destination, rather than just one chance.

dr|z3d sure, circuits will persist for persistent connections, like ssh. but the default duration of a circuit is 10m.

mesh this is why tor can be so fast I think. once a good circuit is found tor can dump as much data as possible over that circuit for however long it takes. but after ten minutes, i2p will kill the tunnel and force renegotiation of a new one

dr|z3d pros and cons to both approaches.

mesh it's an interesting tradeoff. i2p is a very paranoid protocol and really doesn't trust the other nodes. but in an ideal world i2p could allow peers to negotiate long-lived tunnel times (at the cost of less security) and then you could basically implement tor on top of i2p

dr|z3d suggest it to zzz. see what he says :)

mesh he probably wouldn't like it. it would be a lot of work to arguably make i2p less secure hehe

mesh but it would be an interesting research project: increase i2p throughput through negotiating long-lived tunnels and implement tor functionality on top of i2p

dr|z3d there was a research paper published a while back that suggested grading tunnels by latency or throughput to permit more fine-grained control over tunnel selection.

mesh dr|z3d: link?

dr|z3d you'll have to find it :)

dr|z3d maybe it's linked on i2p.net

mesh dr|z3d: I'll look for it. That sounds very interesting.

mesh the problem with the tor approach is that it's very easy to block. Anybody can download a list of tor relays in real time and kill all connections to them at the isp level.

dr|z3d yup, indeed.

dr|z3d the server/client model works for bandwidth, less so for overall diversity.

dr|z3d when most of the world has migrated to fiber, i2p will be an entirely different proposition.

mesh why do you say that?

dr|z3d ?

mesh I mean why will fiber change things?

dr|z3d because you'll be less constrained by individual hops in your tunnels.

mesh oh you mean the quality of tunnels will be better because you'll have less chance of tunneling through a potato

dr|z3d exactly.

dr|z3d and the default b/w allocation can then be ramped up.

mesh yeah fiber will be good. but even if we don't wait for fiber we really just need more people to run i2p routers in datacenters. $10 - $20/mo isn't too bad. imagine if we could triple or quadruple the size of the network (<80,000 nodes) with high quality hosted routers

parabo hey guys

parabo T3s|4, how are you sir?

parabo itsjustme, you awake?

mesh dr|z3d: btw the css override fix doesn't work i2psnark

dr|z3d it should do, assuming you added an override.css file to your snark theme dir.

dr|z3d but if you're running the latest build, you shouldn't need to.

mesh dr|z3d: I'll stick to the stable release for now, but I just wanted to let you know in case you forgot to test that

mesh not sure where i2psnark gets its css but it doesn't seem to respect the override mechanism

dr|z3d each webapp requires a separate override.css file.

mesh ugh

dr|z3d i2p/docs/themes/snark/{themename}/

mesh i2psnark uses a lot of memory. it's impressive in a way.

dr|z3d if you think that's impressive, try Bigly BT, makes snark look conservative.

mesh I truly am curious what it's doing that makes it swallow gigabytes of memory

mesh even for a java app that's impressive

mesh interesting biglybt is also a java app

dr|z3d oh, and you know snark has a debug mode?

mesh dr|z3d: no...

dr|z3d see where your torrent control buttons are? .. follow them down to the footer with your mouse until you expose the hidden debug icon...

mesh hehe, cool. a secret debug mode

rambler ayyo

GalaxyNova hi

rambler How's it going?

GalaxyNova good

GalaxyNova I'm making a postman mail address

GalaxyNova I gave up on trying to selfhost my email

GalaxyNova it's so difficult...

rambler github.com/mailcow is pretty good, but I agree. It can be a big PITA to self-host email, or at least do it in a way where your email will arrive in the inbox of your recipient

rambler I've been wanting to get into some more email stuff, but everytime I start I want to slam my head in a door.

GalaxyNova exactly

GalaxyNova email is just a bunch of hacks on top of decades old protocols

rambler If you're just setting up a you@mail.i2p email, that's dead simple. Should be able to do it from the router console UI I do believe

GalaxyNova yeah that's what I'm doing

mesh it's interesting to consider email over i2p

rambler Feel free to shoot me an email at rambler@mail.i2p to test it

rambler once you get setup

mesh but yeah email is one big hack

rambler I've been wanting to run a clearnet bridge since the existing one may take 8+ hours for the email to be delivered. Not sure what all is going on to cause that, but I think there is a public I2P mail relay, forget the name now

rambler An improvement on that would be a big improvement for email over I2P in general

mesh rambler: how would a bridge work? do people send mail to alice@{destination-hash}.b32.i2p ?

GalaxyNova no there's a specific clearnet address that it's sent from

GalaxyNova i'm pretty sure for postman it's i2pmail.org

mesh I was thinking such a scheme might work but it'd be less than user friendly and possibly prone to abuse

rambler I want to say that if your email was, for example, mesh@mail.i2p then you could use mesh@i2prelay.com or whatever instead

rambler And it'd pipe it to you. Useful for forum registrations and whatnot

rambler Yeah, i2pmail.org is it. Thanks @GalaxyNova

mesh yeah that works... kind of a mail forwarder type scenario

GalaxyNova term99: you should fix the install destination for the freebsd init scripts in i2prouter

rambler Yeah, but it's pretty big delay from having something sent and having something received. But if you're like me you may check your I2P mail once every week or two, ha

GalaxyNova they should go in /usr/local/etc/rc.d

GalaxyNova i barely even check my clearnet mail

GalaxyNova it's riddled with spam

mesh I dunno if b32 i2ps are actually valid email addresses but it'd be cool if ppl on clear net could send mail to something like alice@{xxxb32.i2p}.i2pmail.org and the bridge forwarded that to your email server

rambler No way to filter it? I'm tied to mine, unfortunately (for work, where I am right now actually...). Good spam protection / filtering, plus I just create inbox sieve filters to sort mail in a way that makes sense for me.

GalaxyNova i don't think .i2p is a valid tld

rambler Speaking of work, gotta go! Have a good morning / day / evening / night, all.

mesh GalaxyNova: the tld would still be i2pmail.org. the b32 is just embedded in the email address so that the bridge could forward the email to your mail server running in i2p

parabo dr|z3d

albat hi dr|z3d, hi itsjustme :)

parabo hi albat\

parabo hi albat

albat hi parabo :)

parabo how are you?

albat how are you? :)

albat fine :)

parabo I am good

albat i have too many uploads over retroshare

parabo Have you watched Seinfeld?

albat i own a channel with plenty of tvshows

albat no i didn't

parabo you should watch it

parabo it's a REALLY good 90's comedy

albat ok

parabo it was probably the biggest comedy show of the 80/s90s

albat look : prnt.sc/stSy3h33YGZE

parabo cool

albat prnt.sc/ePeO6nP8AIne

albat i have too many leechers lol

Irc2PGuest40217 If you liked Seinfeld, you should take a look at "curb your enthusiasm"

albat ok

albat hi RN :3

RN aloha albat

albat how are you today?

parabo RN

parabo Curb Your Enthusiasm is great

RN recovering. Had a bunch of things happen including dental surgery and my car caught on fire.

RN yes, I binged the heck out of CYE

albat oh :(

RN all done with the strong painkillers. :(

albat bad :/

RN so that is a good thing

RN pain is under control with over the counter pills, but not feel as good as the narcotics ;)

RN car situation is a work in progress

parabo RN: What kind of car?

parabo I just bought a 1978 sunbird

parabo Wagon

RN rather not disclose details, but a 2012 sedan

RN nothing special, just a utilitarian commuter car

parabo I am getting the Wagon towed to me tomorrow

RN nice

parabo yeah

parabo $800

parabo for a beautiful classic

parabo no accident's but it's got 320,000 km on it

parabo also only 1 driver

RN whhof

parabo gonna fix it up\

scottpedia hello ppl

scottpedia zup guys

term99 hello

scottpedia its been a while

albat hi dudes :)

scottpedia yo albat

RN :)

albat all's fine?

scottpedia yeah

albat i'm soon at washington DC, at the end of the month, to visit my sister, for 3 month

scottpedia alright

scottpedia sounds good

albat she is pregnant, she will have a boy after her girl who is 2.5yo

RN albat, you sure get around. Should we start calling you travelbat?

albat lol

RN travelbator

RN Σ:Đ

albat i will pilot my dektop computer from wahsington

RN remotely?

scottpedia well its not a good time to have kids these days imo...

albat isp at washington DC are lame

albat yes

albat she had her baby girl at 39yo

scottpedia well thats quite unusual