IRCaBot 2.1.0
GPLv3 © acetone, 2021-2022
#ls2
/2023/01/09
@eyedeekay
+R4SAS
+RN
+RN_
+T3s|4
+Xeha
+acetone
+orignal
+weko
Irc2PGuest5995
Irc2PGuest89954
Leopold_
Onn4l7h
Onn4|7h
T3s|4_
aargh2
anon2
eyedeekay_bnc
hk
not_bob_afk
profetikla
shiver_
u5657
x74a6
orignal and with wrong signature ofc?
dr|z3d not too many of those seen, I'm taking a hardline approach when I see them now. permaban.
dr|z3d maybe 2 or 3 knocking around the network.
RN toasters, toasters, toasters everywhere!
orignal I know where they are coming from
orignal i2pd can send only 2 fragments in SessionConfirmed
orignal if RI is longer last part will not be sent
dr|z3d yeah, understood, orignal, RI is being truncated because some clown has configured a stupid number of introducers on X tier router(s).
orignal need to detect this siatuatiomn and raise error
dr|z3d yeah, ideally just spit out an error and shutdown the router.
dr|z3d check the size of the local RI and shutdown if it's too big. ERROR: Too many introducers configured, invalid RouterInfo generated. Shutting down...
dr|z3d 3 excessive introducer routers identified to date, all banlisted here. you want the hashes, zzz, or you don't care?
orignal btw, I remeber I saw SessionConfirmed with more than 2 fragments
orignal my question is why
zzz I think our existing protections are sufficient, and if he's not overflowing the 3072 buffer, it would still be a valid RI and accessible via NTCP2 or outbound SSU2
dr|z3d this is only for RIs that fail verification.
orignal I think they have channged it
orignal the problem that it doesn't go through is SSU2
orignal zzz but where 2+ fragments come from?
orignal i2pd never sends more then 2 in SessionConfirmed
zzz I assume it's only sending 2, so it's truncated, and the sig fails, as you said yesterday
zzz but maybe it's a DSM, not SSU2
orignal no sometimes I see 3rd fragment
orignal in SessionConfirmed
orignal I'm not talking about that guy
orignal I'm asking why Java even send 3 fragments?
zzz if it calculates it's too big, of course. For IPv6 there's 133 bytes of overhead, so for a 1280 MTU that leaves 1147 bytes, so about 2294 bytes for two
zzz that's pretty big but not impossible
orignal then I should implement 2+ I guess
zzz dmLeecZhuY-NEXZ1nZRYOexI8bzzRfrL8rFStCFaUn4= i2pd is 2109 bytes uncompressed right now
zzz if you're going to publish those tunnel build stats you'll need it
zzz turning off SSU 1 saves you a lot of space. 4 addresses with introducers was pretty big
zzz you probably have 2.44.0 routers over two fragments if they're firewalled + ipv6
dr|z3d what's the timeframe for switching off ssu1? mid-year?
zzz see top of prop. 159
zzz btw I see several i2pd routers with compressible padding
zzz some probably stupid prestium
orignal what is prestium?
zzz no persistence, new RI and reseed every time
orignal I thought I did something wrong with copressible padding
zzz looks fine to me, but I hope you tested it ))
orignal I did
orignal you should also see it on zzz.i2p
orignal incoming connections
zzz yeah, haven't looked at leasesets yet
zzz 0) Hi
zzz what's on the list for today?
eyedeekay I can do a short go-i2p update
orignal I think recent high trafiic and number of transit tunnels
zzz ok, you will be 1)
orignal also release
zzz I'll add:
zzz 2) is high traffic
zzz 3) is release
zzz 4) is my symmetric nat state machine proposal
zzz 5) is the D/E congestion caps proposal
dr|z3d 5) perhaps bitcoin progress vis-a-vis i2p/i2pd?
dr|z3d or 6)
zzz ok that's 6)
zzz big list, let's cut it off there
zzz 1) go-i2p
dr|z3d he may be lagged, in which case skip to 2 and return to 1 later.
eyedeekay OK just a brief one, as zzz already knows I had a wild couple of weeks
eyedeekay I added support I2P and bridging support to a pure-Go bittorrent library, and in order to support that, goSam, sam3, i2pkeys, and onramp all got feature releases last week
zzz you still need some index-to-projects page eyedeekay
orignal does it work?
eyedeekay Yes it's great, super fast actually
orignal we tried deluge and it doesn't
eyedeekay Seeds, downloads, and webseeds
zzz so there's some app that wraps the ilb?
eyedeekay just a terminal one now, but github.com/eyedeekay/bttools
eyedeekay It's a fork from the library I adapted
eyedeekay Yeah I think that has to do with libtorrent-rasterbar
eyedeekay I only superficially understand how python-C++ bindings work, not much help there
zzz Vort found the cause
eyedeekay Getting that fixed will be exciting to some users I think
zzz kinda wild it broke 4 years ago
zzz and what about go-i2p, which was the topic? :)
eyedeekay I only just found out that Deluge offered some kind of option for it
eyedeekay go-i2p itself didn't progress as much, Mostly I've been trying to get the obfuscation and de-obfuscation parts of the handshake for NTCP2 right
zzz anything else? we gotta keep it moving, big agenda
eyedeekay Let's call it there, it will take me too long, maybe I'll go to zzz.i2p and do a long-form one this time
eyedeekay I find myself thinking of more to talk about than I expected
dr|z3d ok, 2) high traffic.
zzz ok then, anything else on 1) ?
eyedeekay No thanks
zzz 2) high traffic
orignal the situation looks better now
zzz this is orignal's agenda item, go ahead
dr|z3d still seeing i2pd routers hosting 18K tunnels, orinal?
dr|z3d *orignal
orignal so do we know what cuases it?
zzz my theory remains the same. tunnel build spam
orignal Transit Tunnels: 7241 on mine
orignal twice less
orignal I have another theory
orignal duplicates in SSU2
zzz you have a diffent theory?
orignal about traffic
orignal if same message were resent few times
zzz maybe
dr|z3d we've got 2 separate issues here. a) traffic ie bandwidth usage and b) transit tunnel hikes.
dr|z3d they may or may not be related.
orignal I also think they are different
zzz one thing for sure: i2pd is now a big part of the network. Congestion control is very important
dr|z3d there's a sizeable chunk of transit requests coming from LU routers. that looks suspect.
zzz java i2p congestion control alone can no longer "save" the network by itself
zzz it is now possible for i2pd bugs or poor congestion control to take down the network
zzz not saying that's what happened, but it's clearly possible
zzz the bitcoin guys were creating 290 tunnels at startup
orignal 2.45.1 will contain duplicates drop
orignal will release wednesday
orignal yes and you know I limit nymber if tunnel build requests sent at the time
zzz I boldly predict expl. build success will be over 40% by next Monday
zzz anything else on 2) ?
zzz now my turn for lag...
zzz 3) release
zzz go ahead orignal
orignal we will release 2.45.1 wednesday
dr|z3d I'm seeing a significant drop in tunnel request by throttling/temp banning L/U tier routers.
dr|z3d with a scaled approach, reject first, then temp ban if the router persists, transit tunnels look a lot saner on routers I'm watching.
orignal we seethat 2.45.0 improved th situation a lot
zzz we should release tomorrow, a day late
orignal becuse fixed too many bugs
zzz will see how it goes
zzz anything else on 3) ?
zzz yup, us too
zzz I'm reviewing our diff, forgot about all the bug fixes from early december
zzz tokens and stuff
orignal another thing
zzz it's a really big release for us, diff-size-wise
zzz go ahead
orignal one ukrainian guy showed me DatabaseStore of 47K
orignal with LS1
zzz wow
zzz don't know how, but you can definitely do it with LS2
orignal myabe the network is full of messages like that?
orignal it was LS1
zzz interesting
zzz if I see one I'll let you know
orignal yes and I have add the code to drop if LS exceeds 3K
zzz anything else on 3) ?
zzz 4) nat state machine
zzz mentioned a couple of times briefly in december
orignal yes, I read it
zzz really didn't get it fixed until ~12/26
zzz and I have almost all of it implemented and in the release
zzz so I'd like to put it into the SSU2 spec, but maybe it needs review and testing first
zzz thoughts?
orignal you should add it to the specs
zzz ok, we can always change it later anyway
zzz maybe with some thoughts on "full cone nat"
zzz I'd like to actually test full cone nat and symmetric nat locally, but I don't want to f-up my firewall
zzz maybe with an old firewall behind the outside one
zzz anyway, if you have any edits/corrections/suggestions about it, let me know
orignal use iptables
zzz good idea
orignal to simulate it
zzz or I have an old openwrt box I can put behind my other box
zzz and do iptables on that
zzz anything else on 4) ?
zzz 5) congestion caps
zzz I took the conversation from the other day and wrote it up
orignal what's that?
zzz which is the opposite of what you originally proposed with "high performance" cap, remember?
orignal let me read
zzz so when the conversation ended, I just wrote it up where we left it
orignal so how would you like to publish it?
zzz this would be in the main RI caps, like PfRD
orignal got it
zzz I believe that's what we were discussing, I just wrote it down the way I understood it.
zzz dr|z3d, does the writeup reflect your memory?
orignal so if we see such cap we should try to build a tunnel?
dr|z3d yeah, that's about right, zzz.
zzz no orignal, if you see a cap you should not try to build a tunnel, or at least try less often
zzz they are "congestion" or "go away" caps
dr|z3d if we're going to publish congestion caps, we could do with some method of determining how loaded the cpu cores are.
orignal yes, that's what I reamnt
orignal meant
orignal exclude on
zzz the one thing I added in the writeup is, if it's 'E' but too old, treat it as 'D'
orignal yes, good idea
orignal let's go ahead
dr|z3d you got that in the writeup: "If this RI is older than 15 minutes, treat as 'D'"
zzz right
zzz I'd like to think about it a little more and decide in a week or two. I don't want to make any definite decision until after the relase
zzz and I have more room in my head
orignal btw, if "no transit" I aslwys raise "D"?
zzz if you don't allow any transit?
dr|z3d no transit being currently hosting no tunnels, or rejecting all transit requests?
dr|z3d if the former, D is fine, if the latter, E makes more sense.
zzz if the latter, we need a new letter
orignal yes, I have such option
dr|z3d maybe 'R' for no transit, aka reject.
zzz because E turns into D after 15 minutes
orignal I reject tunnel requests with code 30
zzz R is reachable
zzz you could use code 50 to say go away forever
zzz dr|z3d?
zzz guess he's lagged out
zzz weko, what's your 7) ?
dr|z3d 6) bitcoin.. sorry,
dr|z3d well, we appear to be making progress with the bitcoin team, zzz and orignal have both nudged them in the right direction.
dr|z3d they've updated their i2p documentation, though there's still more work to do there. (20 transit tunnels etc).
dr|z3d orignal, zzz: anything to add?
zzz I acked both the PRs
zzz ok anything else on 6) ?
dr|z3d not from me.
zzz do we have time for 7) ?
dr|z3d 3 minutes.
zzz 7) weko
dr|z3d makes the meeting a round hour.
zzz go ahead
weko Maybe latter
weko I not write yet
zzz ok, next week, if you want
zzz anything else for the meeting?
weko It is suggestion about polling I2NP packets
zzz oh wait. are we on one week or two week cycle?
eyedeekay We were on 2 weeks before the holiday
zzz orignal, one week or two?
zzz yeah but then we had an extra one
zzz ?? you want a meeting next week or in two weeks?
dr|z3d 23rd Jan then.
orignal two weeks
zzz sounds good
zzz happy release week
zzz thanks everybody
weko I suggest add I2NP "polling" message. Tunnel owner send "polling" in tunnel (directly with outbound tunnel or via our another inbound tunnel for inbound tunnels), and every transit router can add some metadata for owner. Then, "polling" will be received directly by router, or received by Endpoint and endpoint sent this packet to our inbound tunnel (we say tunnel(s) in packet)
weko This is useful for:
weko 1) say what out router shutdown. We just need wait 30 sec (if is maximum delay for this messages 30 sec), and in this 30 seconds we add metadata "I am shutdowning...", and then tunnels owner change tunnels for their data without data's streams's lags. It is useful, because we need wait 30 sec (or time what we choose) instead of 10 minutes
weko 2) dynamic change of per-tunnels transit traffic limit. For example, transit router can say "I can transit 150 kbps for now, and then can say " I can transit 200kbps for now"
weko 3) We can use it as tunnels testing (we don't need send more packets without activity with this way).
weko Oh I wrote it
weko We add "I am shutdowning" in every polling message in our transit tunnels*
weko What do you think about this feature?
dr|z3d weko: as I mentioned before, telling 5K routers, or 10K routers that you're about to shutdown is an anonymity risk.
weko dr|z3d: node status is not private info
dr|z3d all I need to do as an attacker is make sure I'm always building tunnels through you, and then I just monitor your shutdowns and the netdb. game over.
weko dr|z3d: we can do same thing now. Just send tunnel requests for getting node status (on/off)
weko dr|z3d: you don't understand my suggestion, I think
dr|z3d there's a reason it takes up to 11 minutes to shutdown. and that reason is predicated on anonymity.
dr|z3d that takes significantly more effort than being told "I'm going offline now".
weko Fact, what you say "I am shutdowning" in transit tunnels (not your tunnels) don't deanonimize you, because router online / offline status already public info
dr|z3d if my router is restarting, there's maybe a 60s window where it's offline. during that period, an attacker can determine I'm offline by probing my port. that's a small window.
dr|z3d I don't currently publish the fact that I'm about to go offline.
weko We say tunnel's owner already public info. But we do this more operative with this.
weko But anyone can check it
weko If you answer, you offline. If not answer, you offline
dr|z3d any proposal has to weigh up the benefits vs the risks. what you're proposing is out of convenience at great risk to the anonymity of the router operator. not worth it.
dr|z3d there's nothing to stop you from aggressively restarting your router without waiting for transit tunnels to expire. the network can handle it.
weko dr|z3d: what risks have publishing public info?
weko dr|z3d: with lags
dr|z3d just read again what I wrote above, weko. there's a reason the same logic's been in play for the last 20 years.