#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

darius must go 4 now

RN I'm going to have to watch closer, but I do believe I've seen connections quit (read error) on ilita, same as they do here or any other I2P facing irc, just different frequenies

zzz I have no cosmic expertise, sorry

orignal RN I'm talking about myself

zzz re: dup tags, you'd have to find some birthday paradox calculator. I don't know if we did that or documented it in the proposal

orignal I never get dsiconnected from there

orignal so can we have duplicate tags?

zzz sure

orignal and how do you resolve it?

zzz it's just some HKDF of something

zzz in your mapping of tag->key, you either keep the old one or take the new one, I don't know if we recommend one or the other in the spec

orignal I use tag and immeditely remove

orignal for incoming

orignal let me see what I do when generate

dr|z3d you're off the hook for now, zzz, he's afk :)

zzz but if you pick one or the other, and you pick wrong, that's only one lost packet, no big deal

zzz that's part of the protocol design, I just don't recall if it's implicit or explicitly documented

orignal btw, what you max num of generated tags

orignal you generate forward

orignal I have 320 for now

zzz same here

zzz e.g. ID: 4/11539 expires in: 9 min with 320+64496 tags remaining

zzz we've generated 320 and there's another 64496 remaining after that

orignal trying to understand what happened

orignal definitly unknown tags on my side

zzz yeah but I'm only sending acks, 10% of what you're sending me, all in the same tagset

zzz tags are 8 bytes so maybe you can find a birthday calculator for 320 / 2^64 but my guess is that's extremely rare

orignal what if a tunnel started malfuctioning and massive loss

orignal more than 320

orignal it's pretty possible

zzz again, I'm only sending acks

zzz ID: 0/10625 acked created: Aug 16, 2024 4:08 PM last used: Aug 16, 2024 4:34 PM expires in: 7 min with 62669 tags remaining

orignal yes but how often?

zzz so in 26 minutes I sent... 2866 acks

zzz 110/minute

zzz so 320 is ~3 minutes

zzz but I request a ratchet ack once a minute

zzz and would have switched tunnels if I didn't get an ack

orignal is it possible that you were not abble to decryot mine?

zzz sure, I'm not doing any logging

orignal say what if I pumped 320 messages to nowhere

zzz but you said you were getting lots of decrypt errors

orignal yes, but it's not necesary about your stream

orignal there were few connections

orignal and I have them often

orignal most likely duplicated packets

orignal zzz seems I have identified the issue

orignal moreTags = ECIESX25519_MIN_NUM_GENERATED_TAGS + (index >> 2); // N/4

orignal where ECIESX25519_MIN_NUM_GENERATED_TAG = 24

orignal simply spekaing I generate too small number of tags at start of a new tagset

orignal how do you calculate number of tags to generate?

darius I'm back and still the same issues, as expressed amid the work ya'll good folk where doing earlier.

darius ihope we can get to the bottom of all these issues, but patience, we'll get there :)

Anomaly i want to change my nick to "Shrek"

darius spare a thought for the ppl being slaughtered in gaza i guess, they really dont have time on there side :((((((

Anomaly oh yea

Anomaly yea that stuff is not good

Anomaly i am cautious of discussing topics like that because it can become political, and i dont think its a good idea to talk about politics.

darius yeah its psychotic, the way they are going about flattening the place, anyone here knnow the latest in the numbered slaughtered and injured, its

Anomaly are you in that area ?

darius you might imply not

Anomaly ok

darius its last i heard the ALL? the twenty something hostipals and clinic were destoryed, terrible drimes

darius crimse

darius have you tried clicking on your name? or right clicking?

Anomaly no, why do you ask ?

darius "Shrek"

Anomaly *** scratches head ***

Anomaly oh like i can just change my name ?

dr|z3d ramble.i2p/f/Politics/5494/the-life-of-yahya-sinwar-the-leader-of-hamas-in-gaza

darius is it as psychotic and genocidal as nteanyahoo? theyve only been oppressed, treated as lesser ppls and had their land stolen for decades, but dunno.

darius oh and that was about to be normalized last year

dr|z3d Read the full article, enlighten yourself. It's a good article, dispassionate.

Anomaly wtf

Anomaly shrek: is that u ?

Anomaly this is my swamp!

Anomaly imagine we have shrek and "Donkey" as a nick

cumlord what are you doing in my swamp??

Anomaly Donkey!

darius the lancet (uk journal) have put out an estimate a few weeks ago that the number of ppl killed at the direction of israel is 186,000 so i was just wondering how that number has changed about a month later.

Anomaly ummm

darius thats not including injured and ppl whove been able to get out, btw, terrible.

darius well done, shrek

darius your transitioning, wonderfully

cumlord like a butterfly

Anomaly lol

Anomaly you guys are very funny

darius it needs a kiss clearly

dr|z3d you didn't read that article, darius. instead of telling us what we already know, inform yourself on what you don't.

Anomaly i think we should fulfill the book of revelation and drop a nuke on isreal

Anomaly im joking btw

Anomaly dont do that

darius its on the reading list, i just wanted to know if numbers had updated, i didnt know everyone knew

snex Nuke em all

darius enuf nihalism its not funny

dr|z3d that is, if you're actually interested in a educating yourself about Gaza.

darius identities are identities, i don't believe all the ppl massacred are/were following that identity, the ppl ive heard from dont say there are many different factions and people in leadership

darius Blinded message

darius also that hamas was ELEVATED by israel

darius the PLO was there and cast aside

darius the UN are pretty clear that israel is conducting crimes against humanity, and need/needed to return stolen land

darius bbl

RN does darius not know they are here?

RN *** checks self and takes drunk elsewhere ***

T3s|4 dr|z3d: thanks for the heads up, I read the entire outstanding piece in The New Yorker :)

zzz orignal, 12 for NSR, 24 for ES #0, 320 for ES > 0

orignal I'm asking the algorithm

orignal so, tagset 0 is range 24-320

orignal tagset 1+ alsways 320. right?

zzz correct

orignal how do you calculate for tagset 0? based on index like I do?

orignal higher index more tags

zzz if (_maxSize > _originalSize) {

zzz // grow from originalSize at N = 0 to

zzz // maxSize at N = 2 * (maxSize - originalSize)

zzz // for typical loss rates, this keeps us at about maxSize,

zzz // but worst case about maxSize * 2

zzz lookAhead = Math.min(_maxSize, _originalSize + (usedTagNumber / 2));

orignal usedTagNumber is index I guess

orignal and you divide by 2 rather than by 4

orignal so you number fo gerenrated tags grow faster

zzz yeah lookahead is how many tags ahead of the tag we just got

zzz so we could have 320 ahead and 320 behind

zzz but if no loss, then just 320 ahead and zero behind

orignal thanks. will fix it

orignal and for next tagset it's 320 from index 0

zzz right. of course these are all just implementation choices, not specified, but I've forgotten so much it would take a lot of time to get my head back in it

zzz I'm lucky I can find the code at all ))

orignal great

orignal anyway it seems it solves that problem with big files

orignal I have forgotten too

orignal it worked well while everything was slow

orignal but new speed enlights old bugs

zzz yes when things are working well it's easier to find bugs

zzz when everything is broken it's very hard

orignal nobody tried to download files of few hndreds megs before

zzz happens in bittorrent all the time

zzz re: collisions, we did talk about it in prop. 144:

zzz A target of 1 in a million (1e-6) session tag collisions is probably sufficient. The probability of dropping a message along the way due to congestion is far higher than that.

orignal and what's the colculsion? new or old?

zzz With 8 byte session tags (64 bits) the session tag space is 1.8e19. The probability of a collision with probability 1e-18 requires 6.1 entries. The probability of a collision with probability 1e-6 requires 6.1e6 (6,100,000) entries. 1.8 million tags of 8 bytes each is about 15 MB total.

zzz 6.1 million active tags is over 3x more than our worst-case estimate of 1.8 million tags. So the probability of collision would be less than one in a million. We therefore conclude that 8 byte session tags are sufficient. This results in a 4x reduction of storage space, in addition to the 2x reduction because transmit tags are not stored. So we will have a 8x reduction in session tag memory usage compared to

zzz ElGamal/AES+SessionTags.

zzz Implementations should, at a minimum, recognize session tag collisions, handle them gracefully, and log or count the number of collisions. While still extremely unlikely, they will be much more likely than they were for ElGamal/AES+SessionTags, and could actually happen.

orignal so print to the log

orignal I use old tag for now

zzz it's going to be extremely rare. The proposal does not make a recommendation on old vs. new:

zzz Therefore, the only concern is session tag collision. It is assumed that implementations will not attempt to handle collisions by trying to decrypt with both sessions; implementations will simply associate the tag with either the previous or new session, and any message received with that tag on the other session will be dropped after the decryption fails.

orignal yes, but still should write to the log

dr|z3d glad you enjoyed it, T3s|4, long read, but worth the effort :)

dr|z3d zzz: can we make publishing a version a hard requirement for RIs? with all the no-version RIs out there right now, it smells like an attack.

dr|z3d also, re specs, if we have something relating to max participating tunnels before throttling in the specs, orignal says he'll implement it. long shot, I know, but I figured I might as well ask :)

zzz re: RI, impls don't need a spec to drop any RI they want. You collaborated with eyedeekay on his recent MR? Please comment there, I nacked it but maybe it can be salvaged

zzz re: throttling,that's also up to the impls, not really appropriate for a spec

dr|z3d re no-version RIs, no collaboration with eyedeekay, I just brought the fact that I was seeing a huge number lately to his attention and he then saw the same behavior on one of his routers.

dr|z3d currently I'm banning them in the participating throttler.

dr|z3d private void handleNoVersion(boolean shouldDisconnect, Hash h, boolean isBanned, String caps, int bantime) {

dr|z3d if (shouldDisconnect) {context.simpleTimer2().addEvent(new Disconnector(h), 60 * 1000);}

dr|z3d if (!isBanned && _log.shouldWarn()) {

dr|z3d _log.warn("Banning Router [" + h.toBase64().substring(0, 6) + "] for " + (bantime / 60000) + "m -> No router version in RouterInfo");

dr|z3d }

dr|z3d context.banlist().banlistRouter(h, " <b>➜</b> No version in RouterInfo", null, null, context.clock().now() + bantime);

dr|z3d }

zzz but are they also missing netid, or is it a mix&match of what's present and not out of {version, netid, caps}? eyedeekay didn't provide any data or analysis in his MR

zzz I'm in the dark because I'm not seeing any of it, perhaps due to local changes

dr|z3d all good questions, not sure since they're all banned and the RIs purged, but give me a moment or 2 and I'll have a list for you to cross reference.

dr|z3d pm'd

dr|z3d regarding transit throttling, perhaps some recommendations/suggestions with an explanatory note on why throttling is good would suffice.. distributing the traffic across the network, better network health in the event a router crashes, that sort of thing? maybe that would nudge orignal in the right direction.

orignal 1 hour cpntiniously. no disconnects

orignal zzz, so trottling is not a requirement

orignal like dr|z3d claims

dr|z3d I never claimed it was a hard requirement, orignal

orignal you meant that non-trottling makes a trouble to whole network

orignal not only my router

orignal it means it must be in the specs

orignal if not it means it's only my business

dr|z3d this is what we're currently discussing :)

orignal fixing ratchets is more important for me

dr|z3d once upon a time I recall zzz suggesting that excessive hosting of transit tunnels was bad for the network, at a time when 13K was deemed to be excessive.

dr|z3d if there were some suggestions in the spec relating to throttling, would you pay any attention to them?

orignal as I said. Once it's in specs we will implement

orignal otherwise it's only our business

dr|z3d so I'm asking, if there were some guidelines/recommendations in the spec instead of a hard requirement, would you heed them?

orignal again all recoomendation must be in specs

orignal to point guys to the source

orignal like that's not my inbvention "to slow down I2P" etc.

dr|z3d ok, so if there were recommendations _in the spec_ you'd take them into consideration. ok, smells like we're getting somewhere :)

zzz I'm saying we have 20+ years of innovation in congestion control and active queue management (AQM), both in I2P and in academia; why constain ourselves to a particular impl in the specs?

dr|z3d I'm not suggesting a specific implementation, I'm suggesting some general recommendations.

orignal because dr|z3d claims it negatively affect the whole network

zzz I don't think convincing me to document something so it will convince i2pd to do what you want makes any sense

orignal if you believe it makes bad impact to the network you should document it

dr|z3d I respectfully disagree. orignal is telling us that in order to keep his users happy, he needs to be able to refer them to the spec if he decides to implement it.

zzz we're all learning together, it's not efficient for me to document everything in canon impl.

zzz and drz changes every single knob and line of code he can find, it's not like he thinks canon impl is perfect either

dr|z3d true that, can't disagree there.

dr|z3d *** chuckles. ***

orignal it's not about implemntation

orignal it's about limits

zzz y'all are well familiar with my recommendations: everything has limits, dropping is good, AQM is good, Westwood+, RFCs for RTT/RTO calcs, etc.

dr|z3d let's try a simple question. 1 router, 40K transit tunnels. good or bad for the network?

orignal we have 100K transit tunnels limit

orignal but dr|z3d is not happy with it ))

dr|z3d *** shakes his fist at orignal, obligingly. ***

orignal my point, if it cvan handle 40K no problem

zzz then you two have a debate, don't make me be the referee or give me work to do ))

dr|z3d we've done the debate. orignal will only consider a different throttling strategy if there's something in the specs.

orignal correct. I don't see any reason to reject a tunnel if I have rources to handle it

dr|z3d something, anything. some basic guidelines..

orignal if you don't like me don't build through me

dr|z3d that's a straw man argument. no relevant to the discussion.

orignal and rememeber

orignal an abversary can do it wotjout trottling

dr|z3d sure they can, and we can limit the percentage of client tunnels a single router can appear in, so we have some mitigations.

zzz drz, if you have recommendations on best practices, write up a white paper. AQM, bufferbloat, the whole shebang

dr|z3d you've obviously forgotten the time you balked at a router with 13K transit tunnels and stated that you might have to implement some sort of stats-based limits, zzz. it was a while back, so you're excused :)

zzz I never said I disagreed with any of your points, maybe I do, maybe not

orignal I always have like 15K transit tunnels at FF

snex You guys need to host webinars for this stuff. Open to the public. We can’t keep doing this thing where only 5 people know how i2p works

orignal ofc much more

dr|z3d if you're content to sit on the fence, zzz, nothing I can do about it.

orignal on the fence? )) I though on the ass ))

zzz if you can't convince somebody by yourself, perhaps you haven't made a compelling argument. I don't feel qualified to make low-level recommendations to i2pd, and when I have, it's often mixed results

dr|z3d there is no convincing orignal unless he sees something in the specs, regardless of the merits of said arguments.

dr|z3d regardless, you could, in general terms, outline your thoughts on why routers hosting a large number of transit tunnels might be bad for the network, or, conversely, state why you think it's a non-issue.

zzz dr|z3d, re: no-version RIs, I sampled some from the list you gave me, every one was fine, with version/caps/netid. Either your detection is buggy, or it's a transient issue, perhaps at i2pd startup

dr|z3d ok, thanks, will take a look at the detection code.

orignal i2pd always sets version

zzz but if eyedeekay saw it too, that would rule out a plus bug

orignal before creting and signing RI

dr|z3d private String getRouterVersion(RouterInfo ri) {

dr|z3d return (ri != null && ri.getVersion() != null) ? ri.getVersion() : "0.0.0";

dr|z3d }

zzz then perhaps a bug in some hack/fork/attack version, not saying a bug in i2pd proper

zzz ^^^ that code is all unnecessary add-on hackery, sorry, RouterInfo.getVersion() already does all that for you and returns non-null always, as documented in the javadocs

dr|z3d ok, thx, will modificate.

darius sounds like i've popped on here at just the right lull in the debate, im pro-distrution and pro-limits and importantly pro-getting back to ppl, bout stuff. i said i would research the swap vs ram space thing with tmpfs...

darius well i;m embarrassed to say there was a man page page for that

darius lol

darius the swap space is used when ram space is exceeded

darius pretty simple idea really

darius Blinded message

darius now on the topic of blocking font hinting in the browser for the rss button svg alterative

darius i descided to take a completely different tack

darius div borders....

darius i think ppl will love this

darius cake.i2p/file/lhjzncGzPh_GVJvgj4cAaPUnUvVHqhpEy6TZHP3hb_h81QYwOcIS/rss-using-div-corners.html

darius i think especially dr|z3d will find this fun, and i half expect him to runaway with the little /* TODO */ in the css :P but maybe not

zzz re: thoughts on high part. tunnel count, I'm not going to write your whitepaper for you, but you may wish to explore issues with failure points / consequences if the router goes down, and other possible threat models/attacks/deanon ideas

darius ^

dr|z3d don't expect anything, throstle, you'll avoid disappointment that way :)

zzz the specs are things we agree on, they aren't a mechanism to convince somebody of something, that turns everything upside down

darius yes, I will enjoy doing to todo when i'm less about to acktwolly die

darius sounds bout right zzz

darius anyway dont be scared to offer feedback to the button, i already know that snex will think its a waste of time, but yea. must go and less die

not_bob Mail continues to not work all the time. But, at least it is working some of the time now!

snex i should make something that fwds i2pmail to my regular email

snex like maybe encrypt the entire message first then send it? dunno