R4SAS
lol
orignal
what?
orignal
now I understand my logic
orignal
ipv6 always required mtu
orignal
that's why I have to detect it
orignal
while ipv4 always had a default value
R4SAS
about MTU for v4. like I said, we must check it too because here can be situations where VPN used
R4SAS
not always
R4SAS
VPN will make you use 1280
orignal
but now I know why I didn't do it before
R4SAS
and why "always"?
R4SAS
it is not required if you have native IPv6
R4SAS
because ISP will provide address same wasy as IPv4 when you dialing PPPoE with MTU 1480 or 1500 if IPoE is used
orignal
<zzz> ipv4 is mod 16 = 12, with a default 1484
orignal
that's why I didn't publish
orignal
zzz, what temination reason fo you send if I2NP messaged were not acked?
zzz
14
zzz
any results of investigation into ImQC and k8vh?
orignal
they are not ours
zzz
do you know who? they updated - ImQC is now 1480, k8vh is now 1420... why?
orignal
I have changed the code to use mtu=1480 for HE
orignal
because my code change
zzz
the question from yesterday was, is it a MTU detection bug or a PMTU issue?
orignal
we detect MTU propely on windowns
orignal
that's what we can say
zzz
ok I see you just hardcoded it like I was doing for HE
zzz
is range48 always 1420?
zzz
*route48
orignal
no, but they wireguard is
orignal
better to set it lower than higher
zzz
then why not set it lower for HE? they don't support wireguard?
orignal
looks like every wreguard uses 1420
orignal
no they don't
zzz
ok
orignal
only gre
orignal
hence route48 has lower requirements to your network
orignal
but you need modern kernel
orignal
that supports it
orignal
deavmi uses some german broker for wireguard and also 1420
orignal
will ask him
zzz
you're not doing ipv4 mtu right
zzz
example: y7hn
zzz
SSU: cost: 9 caps: BC host: 213.252.246.27 key: rf8kzcuTCDnMMFGLPhPPa2DQ2yt3nw2lQAwHke-diLQ= mtu: 1472 port: 30673
zzz
must be mod 16 = 12. 1468
orignal
look at 10 lines above
orignal
it's for ipv6 only
orignal
I haven't implemented ipv4 yet
orignal
if (host.is_v6 ()
orignal
where did this example come from idk
orignal
shouldn't be
orignal
I mean I shouldn't publish mtu for SSU1 as for now
orignal
ipv4 SSU1
zzz
y7hnzN4SInr7JdqskYw-rbhkwabFofeGctFGfte-o5Q= IPv4 SSU1
orignal
maybe it's a fork?
zzz
just reporting what I see
orignal
because i2pd never had a code to publish ipv4 mtu
orignal
do you see more?
zzz
no
orignal
maybe it's someone's fork
orignal
I thought we are not supposed to publish mtu for ipv4 before yesterday
zzz
is xZ9n 1500?
orignal
yes
orignal
but why?
zzz
still seeing some failures with it
zzz
not too many
orignal
let me rebuild it
orignal
I haven't updated it for few days
orignal
MTU:1500
orignal
10 SSU2 sessions right now
orignal
2RRY size is 1514 and I;m connected to you thorugh ipv4
zzz
ok, restarting now
orignal
I meant you are able to handle it's SessionConfirmed
orignal
I will rebuild xZ9n in 30 minutes
orignal
a little busy now
zzz
ok
orignal
hmm deavi told me strange this
orignal
his ipv6 tunnel has proefix 2a04:5b81
orignal
with mtu 1420
orignal
but it's not in the list
orignal
updated xZ9N
zzz
ImQC HE addr still failing, advertising 1480, I'm forcing to 1472
orignal
why 1472?
orignal
if it's SSU2
orignal
how would it make a difference?
orignal
if I understand 1472 is for 1480
zzz
I'm just forcing all HE to 1472 max, whether SSU 1 or 2, just for simplicity
orignal
that's what I did in SSU1
orignal
do you want me to set max mtu to 1472?
orignal
instead 1500
zzz
for all SSU2? no
zzz
just trying to figure out if we should limit session confirmed to 1280, and start data phase at 1280
zzz
and then ramp up
orignal
maybe
orignal
how about others?
zzz
at least 90% of my failures are with i2pd still
orignal
it might not be just mtu but something else
zzz
the i2pd IPv4 with a problem is sQ8dVd and ~GIB3b
zzz
but maybe they don't have fragment fix
zzz
if it is PMTU then I'd expect to see more java fails
orignal
maybe Java users don't like HE
zzz
maybe :)
orignal
but you can check by IP
zzz
last 4 hours:
zzz
5 sQ8dVd
zzz
5 ImQCa~
zzz
1 YXEAXl
zzz
1 ~GIB3b
zzz
1 CEFnjX
zzz
1 bAU~6X
orignal
no mine in the list
orignal
let me make 2RRY floodfiil
zzz
I see lots of java HE
orignal
and what mtu do they publish?
zzz
mostly 1472, one is 1280
orignal
I guess for SSU1
orignal
I also made some fix about incorrect buffer size for incoming packets
orignal
that's why I'm asking about 2RRY, xZ9N , MNcW and cdoF
orignal
they all run lastet code