IRCaBot 2.1.0
GPLv3 © acetone, 2021-2022
#ls2
/2023/01/24
@eyedeekay
+R4SAS
+RN
+RN_
+T3s|4
+Xeha
+acetone
+orignal
+weko
Irc2PGuest89954
Leopold_
Onn4l7h
Onn4|7h
T3s|4_
aargh2
anon2
cancername
eyedeekay_bnc
hk
not_bob_afk
profetikla
shiver_
u5657
x74a6
zzz -Akx now (occasionally) publishing D/E
zzz dr|z3d, postman reports irc is running on i2p+, multihomed, please investigate the ls issue
zzz I've asked if he has 'optimize for multihoming' checked and will pass along the answer when I get it
dr|z3d ok zzz, thanks
zzz I'm still thinking about how it could happen but low down my list for now
dr|z3d I had a thought re the assigned letters for the congestion.
dr|z3d letters/flags
dr|z3d ideally they should be consecutive and not overlap existing letters, I'm thinking f specifically.
zzz maybe an emoji instead?
dr|z3d CDE I'm thinking.
dr|z3d very funny.
zzz Caps: PfR🍆
dr|z3d c, capable, d, degraded, e, exceeded, or something.
dr|z3d lol. I KNOW you're joking.
dr|z3d nothing to prevent a visual representation of the state in the console, notwithstanding.
zzz C is used in SSU caps which is a different namespace but still best avoided
dr|z3d I was just thinking C, we have B as well.
dr|z3d A, accepting, D, degraded (partial), E, exceeded, encumbered, evade.
zzz caps = caps.replace("G", "🔥");
zzz we don't need an "OK" cap, that's the default
dr|z3d I know it's not consecutive, but it doesn't overlap f.
dr|z3d what it does do is roughly follow high school marking schema. so sort of intuitive.
dr|z3d so we only need 2 then?
dr|z3d had to zoom to see that emoji. *chuckle*
dr|z3d B, blocking some, C, congested?
dr|z3d sorry, no C.
dr|z3d *slaps self*
dr|z3d so we could still go with D and E for consecutive.
zzz still need G for no-tunnels-ever
dr|z3d so we need 3 then?
zzz yes
dr|z3d can we mix and match letters and numbers? if so, no tunnels forever, "0" ?
zzz putting up the proposals now, with D/E/G, we still have plenty of time to change it
dr|z3d ok, maybe I'm just bikeshedding and it's not that important.
zzz you're definitely bikeshedding but we can still change our minds, bring it up at the next meeting
dr|z3d G kinda makes sense I guess, it's a counterbalance to f which is implies (though doesn't necessarily mean) high bandwidth.
dr|z3d anyways, I'll stop now.
zzz proposals will be up at bottom of the hour
zzz postman reports 'optimize' option is enabled on both mulithomes
dr|z3d what does optimize do exactly?
zzz forces bundling LS in garlic, so client-server association will be sticky
dr|z3d ok, thanks
dr|z3d and when you say sticky, sticky as long as the http client tunnel persists, or just until the lease expires, even if a newer one arrives?
dr|z3d for example, say I hit a few floodfills for a leaseset concurrently, get one, and then a newer one comes in a few millseconds or so later?
zzz this is in the garlic
dr|z3d so the first one wins in essence until the lease expires?
zzz huh?
dr|z3d ok, maybe I misunderstood. if the leaseset's in the garlic, that that as long as I have a connection to the server, I always have the leaseset?
dr|z3d *that means
zzz right
dr|z3d ok. so what happens for non-persistent connections? I have a leaseset to the server while I'm talking to the server, and then I drop the connection.. I'll potentially get another multihome after that leaseset has expired. presumably.
zzz right
dr|z3d so on that basis, assuming I'm a single instance b32, there's no actual downsize in configured it as multihomed?
dr|z3d *downside
zzz bandwidth
dr|z3d bandwidth, ok.
dr|z3d so I'm trying to wrap my head around how you can acquire two differing leasesets and have them both stored in the netdb. that's definitely a head scratcher. ordinarily the newer one should push the older one out, no?
zzz 2 different dests (subsessions), DSA and Ed
zzz that's the orignal complaint
dr|z3d all I can think of is that i2pd isn't handling them in the same way java i2p is, when there's DSA and Ed.
dr|z3d java just ignores DSA, right?
zzz no. I reproduced it from here
dr|z3d so what _should_ happen when a dest publishes both DSA and Ed?
zzz whatever flavor client connects to, that's the LS that should get bundled back
dr|z3d ok, well if you're stumped, color me stumped++
dr|z3d this might be a bit of a stretch, but let's say there are a cluster of hostile floodfills out there trying various attacks to frustate the acquisition of leasesets.. any way to test for that? and could could they in theory cause the issue under discussion?
zzz the way to get unstumped is reproduce it
zzz assuming you don't want to set up an IRC server, standard tunnels with netcat on both ends would do
zzz then find the problem, or retest with canon
zzz if the problem is on your side, you'd have to audit half your codebase to look for some rev you missed, or mis-merged, in the last 10 years, or some other borkage
zzz or, at that point, give up and tell postman to switch to canon
zzz bitcoin update posted to their issue
zzz proposals are live
orignal it publishes both I guess
orignal but if I connect to EdDSA one I expect EdDSA LeaseSet update rather than DSA
orignal dr|z3d DSA leaseset comes from connection not from floodfill
dr|z3d orignal: ok, hmm, thanks.