~dr|z3d
@RN
@RN_
@StormyCloud
@T3s|4
@eyedeekay
@not_bob_afk
@orignal
@postman
@zzz
%Liorar
+FreefallHeavens
+Xeha
+acetone
+bak83
+cancername
+cumlord
+hk
+scottpedia
+uop23ip
An0nm0n
Arch
Danny
DeltaOreo
Irc2PGuest11124
Irc2PGuest50774
Irc2PGuest67483
Irc2PGuest74564
Leopold_
Meow
Nausicaa
Onn4l7h
Onn4|7h
anon4
anu
boonst
enoxa
mareki2pb
onon_1
plap
poriori_
qend-irc2p
shiver_
simprelay
solidx66
u5657_
xeiaso
> there are some subtleties which affect the scenario you described that makes it probably not completely effective
xeiaso
Is there a channel where this is discussed? I'm curious what the subtleties are.
eyedeekay
We discussed it privately when the solution was devised, but we won't be discussing the precise details until later next week.
not_bob
Greetings!
eyedeekay
Oh right I was supposed to ping you not_bob!
eyedeekay
Totally slipped my mind
not_bob
Yes, but my client keeps dying.
not_bob
Something about network attacks...
not_bob
Anwyay, comments on my message?
not_bob
And, I am well ware of the extra network overhead.
eyedeekay
So, re: random +1 hops on client tunnels by default, I think, can be OK, *if* they are new client tunnels
not_bob
Even more so right now.
not_bob
So, not for stock?
not_bob
Also, in my testing +1 random really only means 20-30% of the time you get an extra random hop.
eyedeekay
But I don't think we should "migrate" client tunnels so that existing fixed-length tunnels are now doing +0/1 random hops
not_bob
Since longer tunnels are harder to build.
not_bob
*nod*
eyedeekay
I'm considering it for stock, just not for existing tunnels
eyedeekay
*clients
not_bob
I'm cool with that.
not_bob
Thank you for looking at it.
eyedeekay
Yeah 4 or less is the most reasonable performance
xeiaso
later next week as in the next meeting on the 22nd?
dr|z3d
if you're going to introduce variance, 2 default hops +2 may be a good happy medium. more variance, less chance at guessing hop length, and faster as well as slower tunnels.
not_bob
I agree with dr|z3d
not_bob
Also, in my testing setting it to +2 random causes roughly half the tunnels to be +1
not_bob
And that is really the desired outcome.
not_bob
Once in a while a 5 hop tunnel will happen.
xeiaso
that's because the longer hop tunnels have a higher chance of failing
eyedeekay
sounds right
eyedeekay
xeiaso possibly sooner, but yes at least by then
not_bob
xeiaso: Exactly.
dr|z3d
that's only if you're starting with 3 hops, not_bob. 2 +2 and you'll have 4 max, 2 min.
xeiaso
IIRC someone complained of i2p+ building really short tunnels when the TCSR was really low a couple of months ago
not_bob
dr|z3d: Yes, I had assumed we start with 3 default.
dr|z3d
I2P+ isn't doing anything special there, and the complainant probably misinterpreted their logs.
dr|z3d
tunnel hops aren't decreased on repeat failure, tunnel count is.
not_bob
xeiaso: I have used varius tunnel lenghts and have never had issues with short tunnels.
dr|z3d
not_bob: if random is default, I'm suggesting 2 default with a variance of +2.
dr|z3d
5 hop tunnels, no. just no.
not_bob
But, isn't 3 default now?
not_bob
We don't really want to go below 3.
dr|z3d
4 hop tunnels increases overall network overheads and local lag enough as it is. I'm not even sure it's a good idea by default, though a global tunnel option to introduce randomness to all server/client/both tunnel groups might be useful.
not_bob
Right, and I do understand that.
dr|z3d
no, you mean you'd prefer it not to go below 3. but that's you. there's no global preference, and introducing lag globally is therefore contentious.
not_bob
Isn't the default 3 hops?
dr|z3d
it is currently.
dr|z3d
what I'm suggesting is 2 with a +2 variance, if randomness by default is being considered.
not_bob
Would that cause any privacy issues?
dr|z3d
no extra overhead, and as much chance of getting _faster_ tunnels as slower tunnels, and less predictability in the actual tunnel length.
not_bob
I do see what you are getting at.
dr|z3d
arguably it's safer than 3 +1.
not_bob
HOw so?
dr|z3d
> less predictability in the actual tunnel length.
not_bob
You make a fair point there.
not_bob
And it would not really change the overall load on the network.
not_bob
In most cases a 2 hop tunnel is a fine since there are also hops on the other end of the endpoint.
xeiaso
if 2 hop tunnels are fine, why is the default 3 and not 2?
not_bob
But, with stormycloud being 0 hop, I'm not sure how I feel about 2 hops to a 0 hop.
dr|z3d
2 hop tunnels are arguably easier to sybil, but given the bidirectional nature of tunnels, 2 +2 would make it harder to tap inbound/outbound with confidence.
not_bob
It's also good to consider that tunnels expire after ten min.
not_bob
So, the number of hops will change often for a single tunnel.