@eyedeekay
+RN
+acetone
+weko
Hidenet1
Irc2PGuest33877
Minogami
Onn4l7h
Onn4|7h
ProRu
R4SAS
anon
anon3
eyedeekay_bnc
j6
not_bob_afk
orignal_
profetikla
qend-irc2p
x74a6
yourtrueself_
orignal
I'm talking about second
zlatinb
some more data after 36 hours of uptime. Exploratory:
zlatinb
samples: 1202
zlatinb
0% 39ms
zlatinb
25% 280ms
zlatinb
50% 437ms
zlatinb
75% 635ms
zlatinb
95% 1818ms
zlatinb
99% 6497ms
zlatinb
100% 25815ms
zlatinb
Client:
zlatinb
samples: 2600
zlatinb
0% 9ms
zlatinb
25% 189ms
zlatinb
50% 330ms
zlatinb
75% 519ms
zlatinb
95% 1357ms
zlatinb
99% 5284ms
zlatinb
100% 48280ms
zzz
thanks zlatinb
zzz
you have a proposal based on those results?
zlatinb
I'm leaning towrads System.siSlow() 7000 : 5000
zlatinb
but i haven't benchmarked System.isSlow()
zzz
I was thinking timeout = isslow ? 10000 : 5000; LOOP_TIME = isslow ? 1000 : 800 (to meet the constraint in the comment)
zlatinb
that's fine too I think
zlatinb
at 5000ms you're cutting off a bit over 1% of the tunnels that would otherwise succeed
zlatinb
on a separate note, anyone running linux on a mac M1 would innacurately be branded isslow
zlatinb
but that's a niche market for now
zzz
yup, I'd like to be conservative becaue we really don't have much visibility into what the numbers would be on android, or a new router, or some part of the world with crappy internet
zzz
once ElG is gone, we really don't need jbigi, and can take that out of the isSlow equation
zzz
firing up LXC for first time in 8 months
zzz
trying to get them all updated and talking to each other
zzz
and remembering how to do anything at all
zzz
will then work on NTCP-only
zzz
and then SSU2
zzz
zlatinb, did you have any comments on the TF draft for me to pass along?
zlatinb
I received a copy of it too, no comments from me
zzz
heh, didn't even notice him misspelling your name in some places? I will tell him that...
zlatinb
I'm sure he proof-reads things before publishing
zzz
Lance is really the star of the article. I'm stuck in the middle with just a couple quotes
zzz
which is fine