IRCaBot 2.1.0
GPLv3 © acetone, 2021-2022
#saltr
/2022/04/01
~dr|z3d
@RN
@StormyCloud
@T3s|4
@T3s|4_
@eyedeekay
@not_bob_afk
@orignal
@zzz
%Liorar
+Gid_
+Minogami
+Unbur
+acetone
+profetikla
+snex
+uop23ip
+weko
An0nm0n
DeltaOreo
FreefallHeavens
Hidenet
Irc2PGuest77041
Irc2PGuest9172
Irc2PGuest99412
Leopold
Nausicaa_
Onn4l7h
Onn4|7h
ProRu
anon2
anon3
anu
cheddah
itsjustme_
limak
poriori
qend-irc2p
skank
mesh dr|z3d: so yeah performance is much, much better if you set MaxGCPauseMillis to 500
mesh after 12 hours of uptime I'm seeing much less memory usage and better download rates
mesh so MaxGCPauseMillis the smaller the better with I2P+
mesh it makes you wonder if it would be even better off with the default MaxPauseGCMillis=200
dr|z3d that's good, mesh, but it's probably a little early to draw hard conclusions about download speeds. the longer your router's up, the better the overall performance will be on account of peer profiling.
dr|z3d when you're seeing 99% green ticks on /profiles your router's doing well.
dr|z3d if you want to test download speeds and help I2P+ at the same time, grab all the torrents listed on tracker2.postman.i2p/index.php?view=TPoolDetail&id=1256 and download them all at the same time in i2psnark. there are some very fast servers there that should deliver whatever your connection is capable of, tunnels permitting.
dr|z3d I've seen torrents from that pool download at over 2MB/s
dr|z3d also, if you want a download graph overlay on your i2psnark screenlogs, add the i2psnark inbound graph on /configstats
dr|z3d it's a non-intrusive visual that gives you a rough idea of what your download connection's doing.
mesh thanks for the tips
mesh that's interesting, BiglyBT is Azureus
mesh it's still going after all these years
dr|z3d you can also improve performance by reducing the timeout for peer tests that determines whether a peer gets a green tick on /profiles, based upon the overall average.
dr|z3d bigly is a fork of azureus by the main dev.
dr|z3d routerconsole.showPeerTestAvg=true and routerconsole.advanced=true in your router.config file will show you the average peer latency in the sidebar.
dr|z3d router.peerTestTimeout={value in milliseconds} will configure how long a peer test should take before a peer is determined to be laggy. default is 1s, or 1000ms.
mesh dr|z3d: what's a good value for peerTestTimeout? 500?
dr|z3d depends on what your average is.
mesh ah good point. I tend to just cut numbers in half but probably a good idea to first gather some data.
dr|z3d you'll see average / average fail.. take the first value and round it up.
dr|z3d you can't see it too low, in any event, it'll use the average value if you do.
dr|z3d I generally add another 100ms to whatever the sidebar's reporting and round up to the nearest 100, so if you're seeing an average of 450, round up to 600.
dr|z3d any router that fails to complete the tests in the time you configure will be marked with a red X on /profiles and demoted from the fast tier. keep that page open and it'll autorefresh.
mesh the latest spring exploit btw is yet another example of why http is shit btw
T3s|4 dr|z3d: I don't know what changed that broke my i2p, but I suspect it was somewhere afaict between -9+ and -11+. I can tell you for a fact that after d/ling -11+, and restarting, that breaks my i2p completely. That means I cannot connect to i2pirc, nor to zzz.i2p, postman's tracker or skank.i2p, et al. The only solution I know works so far is to move both ~/i2p and ~/.i2p to
T3s|4 something else, then do a clean install of 1.7.0+ from i2pplus.github.io/#download. I've now seen this identical behavior on both of my i2p laptops.
dr|z3d T3s|4: yikes. anything in the logs?
T3s|4 nothing in the new -0+ log of concern, and unfortunately, I've already nuked the old i2p and .i2p folders
dr|z3d T3s|4: ouch. pretty hard to troubleshoot no logs :|
T3s|4 dr|z3d: ofc
dr|z3d if it's any consolation (it isn't), I haven't seen any issues here with the latest builds.
T3s|4 I am grabbing your latest now; if I have to revert again to -0+ again, I will, and will let you know
dr|z3d ok, keep an eye on the logs, if there's anything preventing the router from functioning, should be entries there.
genka dr|z3d: seems like i set up cropping jpg and png in /img/ via github.com/willnorris/imageproxy
genka libr7vbp7i3lsvqwy5aqgql44uxaufawi5yrloh3x7sezubsoaba.b32.i2p but most likely my i2p router isnt integrated enough
genka and i mostly dont know about peer profiling and peer selection in java router and i2pd
genka so idk
dr|z3d I think I looked at that one before, genka. looks like it's got potential.
dr|z3d let's have a look, loading...
mesh genka: I can see that website
mesh very cool it appears to be a copy of reddit
mesh I wonder how long it takes for new Destinations to propagate around the network and become visible
dr|z3d almost instant.
dr|z3d leasesets are published to floodfills, client queries floodfills, if floodfill doesn't have leaseset, client queries some more.
genka is 5 outbound and 5 inbound tunnels enough, guys?
genka 5/5 is default in i2pd
mesh genka: 3 tunnels and 3 hops is more than enough in my experience
genka for server tunnel
dr|z3d depends on how much usage your server gets.
mesh I guess the more tunnels the better in terms of high bandwidth but there's probably diminishing returns
dr|z3d each tunnel can theoretically handle 8MB/s
mesh dr|z3d: you hit that website?
mesh hehe my buggy code just collapsed. but impressive that time it took about 3 minutes
dr|z3d not loading just yet.
mesh dr|z3d: it won't load, you just crashed the server hehe
mesh it seems to take a while for socket disconnects to propagate. I wouldn't be surprised if your browser still says loading
mesh well my stupid, buggy code crashed the server, but I saw your requests
dr|z3d I doubt you did
dr|z3d not mine, anyways.
dr|z3d I haven't been able to acquire a leaseset yet.
mesh genka: you?
dr|z3d now it's just unreachable, so I have a leaseset.
mesh dr|z3d: took about 10 minutes then
mesh not bad
genka mesh: trying now
dr|z3d less than 10 minutes, more like 3.
mesh dr|z3d: well I published at 21:44
dr|z3d T3s|4_: any joy?
genka nothing yet :/
genka just loading
dr|z3d mesh: for future reference, you should cite any times in UTC to avoid giving away hint about your location.
mesh dr|z3d: good point
dr|z3d you can add a UTC clock to I2P+'s sidebar.
mesh I'm seeing requests funnily enough Received from client: GET / HTTP/1.1
mesh but hehe it doesn't seem to be genka or dr|z3d
genka mesh: nothing. i can see that stream is opened but nothing appears in browser
mesh genka: yeah the server (thread) crashes
mesh I was just curious about the time from when I published the Destination to the time you could access it
dr|z3d sometimes it's more or less instant.
mesh according to zzz it's worse for hidden mode routers
mesh I've seen anywhere from 5-10 minutes though just now it took 4
mesh it's not too bad, but I think it's something my users will have to keep in mind
mesh dr|z3d: I'm curious how you knew when you acquired a LeaseSet
dr|z3d proxy error will change.
dr|z3d from destination leaseset not found to "failed to connect".
mesh dr|z3d: from what to what?
mesh dr|z3d: when you say proxy error you're talking about the message you received from I2PTunnel
mesh interesting
genka so, i need to register domain for libreddit. what registry services are exist except reg.i2p?
genka dr|z3d, mesh
genka what registry services are exist except reg.i2p?
dr|z3d stas.i2p, reg.i2p, those are the main 2.
dr|z3d there's also dns.chudo.i2p
dr|z3d (if it's up)
genka is my libreddit works nicely? maybe i have to do something before domain registration
genka dr|z3d thanks
dr|z3d seems to be working fine, genka.
mesh reg.i2p is interesting
dr|z3d you might want to tell nginx to use the immutable cache control tag on images etc, genka.
genka dr|z3d what about image cropping? not too much?
genka it's set to 250px of width now
genka and proportional height
dr|z3d I'd personally favor rescaling over cropping.
dr|z3d doesn't look too bad, in any event.
dr|z3d blurry, but acceptable.
dr|z3d I think you're rescaling instead of cropping, anyways, no?
genka ah yeah, my bad. resizing, not cropping
mesh but alas http will never be safe
mesh somebody should do an analysis of how easy it is to finger print browsers and track people even on the i2p "dark web"
dr|z3d without javascript, not easy. and if you're browsing i2p, mostly you should have js disabled.
genka dr|z3d so, only thumbnails is rescaled on teddit.i2p, right?
genka libreddit dont have thumbnails and use original pictures
dr|z3d genka: no, both thumbnails and images are treated.
dr|z3d thumbnails are scaled and cropped, images are scaled. all are then optimized and recompressed where necessary.
genka so every picture is rescaled on my libreddit
genka even original
genka dr|z3d oh, so it has better quality than mine
genka i see
dr|z3d no surprise there, you're scaling up from 250px width. probably a bit too small.
genka changed to 512x width
genka 512px*
genka seems nice
dr|z3d much better. are you doing any recompression?
genka hmm...no? just changing jpeg quality to 60 and width to 512px
genka isnt jpg and png already compressed?
dr|z3d well, jpeg quality is compression.
dr|z3d png is a bit different, the default compression doesn't necessarily give you optimal filesize.
dr|z3d have a look at pngcrush and pngoptim and test out some files, see what sort of reduction in size you can get. then you might want to work those into your pipeline. or not.
dr|z3d optipng not pngoptim, sorry.
genka dr|z3d will try a little bit later, thanks
dr|z3d but definitely set a long max-age and immutable on your images, at the very least, via nginx headers.
dr|z3d that way once the client has them in their cache, they won't be requested again, speeding up page loads and reducing server load.
genka dr|z3d: done
RN mesh, how is your research on how easy it is to finger print browsers and track people even on the i2p darkweb going?
RN testing, testing, is this thing on?
T3s|4 dr|z3d: upgraded to -11+, which broke i2p (again). Now running -0+. Here are the log entries while -11+ was running. ix.io/3U1f Lots of kamedlia issues afaict
genka registered libreddit.i2p at reg.i2p
dr|z3d T3s|4: thanks, let me have a look at that.
dr|z3d T3s|4: looks like you unearthed a null pointer exception. I've hopefully fixed it in the latest build.
dr|z3d Also of note in the logs: Configured for 25.60MiBps share bandwidth but only 384.00MiB available memory.
dr|z3d > Recommend increasing wrapper.java.maxmemory in /home/user/i2p/wrapper.config to at least 563 (MB) if the actual share bandwidth exceeds 8.00MiBps.
mesh you're probably better off not setting a max memory
RN thanks for the opinion
RN mesh, how is your research on how easy it is to finger print browsers and track people even on the i2p darkweb going?
mesh you'll get better performance without a max memory, but if you do consider setting a softmaxheapsize, I think
T3s|4 dr|z3d: yep - noted, I keep changing the wrapper.config memory values - but probably forgot to do that last time
dr|z3d T3s|4: yeah, if you reinstalled, then it's easy to forget, not that 384M will give you much of a problem for normal usage.
dr|z3d softheapsize looks interesting, mesh.
mesh RN: it's in the backlog
RN I was asking how your research is going?
RN mesh, you realize "somebody should.." translates to "I volunteer to..." don't you?
mesh RN: not really
dr|z3d yes really. somebody should is a really annoying phrase most of the time.
RN now you know. :)
dr|z3d any armchair pundit can volunteer other people to do work. the world's full of them. we need doers :)
mesh I think people should be able to propose ideas without necessarily volunteering
mesh we already know browsers on the dark web are a huge security risk
RN the internet is a huge security risk
mesh we already know thatthe FBI has penetrated tor web browser several times
mesh RN: sockets can be made very secure
dr|z3d by all means suggest ideas, but "somebody should" is, nonetheless, a tedious phrase.
RN security levels are subjective
mesh the problem is http combined with the monstrosity that is the modern web browser
mesh that has an attack surface larger than some operating systems
RN I agree that is A problem
RN it is easy to point fingers, but doing something about it is the challenge
mesh which leads to the actual question... do people have some ideas on what the actual best "block size" is for i2p?
mesh I remember zzz saying something like too small is bad, the bigger the better?
RN is sounds like you answered your own question. zzz knows a little about I2P and I generally defer to what he reccomends.
RN best is subjective, so your best is not the same as mine
mesh yeah. I need to find an irc client that stores logs and lets me search them
dr|z3d like hexchat.
mesh Very frustrating to ask a question, get an answer, and then forget the answer.
dr|z3d hexchat with logging to files and a big screen buffer and you're good.
mesh yeah
RN also keeping a text file for notes with Q/A pairs is helpful so you don't have to start grepping through logs first
dr|z3d mesh: so when's your blog live?
genka dr|z3d: how are services in the topic of that channel is connected to i2p+ btw
genka and what is saltr