IRCaBot 2.1.0
GPLv3 © acetone, 2021-2022
#saltr
/2025/01/19
dr|z3d > web100 was the NDT protocol used by M-Lab until November 2019. As a part of M-Lab’s platform upgrade, ndt-server replaced the now deprecated web100-based NDT server. The new server provides the ndt5 protocol for backward compatibility with current clients, and will add the ndt7 protocol for future client use.
dr|z3d web100 is the protocol referring to data collected by the original NDT server
dr|z3d Relied on the web100 kernel module for TCP statistics
dr|z3d Collected using the original version of NDT server
dr|z3d Used the Reno TCP congestion control algorithm
dr|z3d Ran from 2009-02-18 to 2019-11-20
dr|z3d ndt5 is an NDT protocol designed to be backward compatible with legacy NDT clients
dr|z3d Relies on tcp-info for TCP statistics
dr|z3d Collected using ndt-server, which follows the legacy NDT protocol to support existing NDT clients that use it
dr|z3d Uses the Cubic TCP congestion control algorithm
dr|z3d Started 2019-07-18 and continues to present.
dr|z3d ndt7 is an NDT protocol that uses TCP BBR where available, operates on standard HTTP(S) ports (80, 443)
dr|z3d Relies on tcp-info for TCP statistics
dr|z3d Collected using ndt-server
dr|z3d Uses the BBR TCP congestion control algorithm, falling back to Cubic when BBR is not available on the client side
dr|z3d Client supports an opt-in data transfer limit
dr|z3d Started 2020-02-18 and continues to present.
dr|z3d maybe this would help with non-responding servers, zzz, dunno: measurementlab.net/develop/locate-v2
dr|z3d also, the simplest upgrade path to their latest protocol may be this: github.com/m-lab/ndt7-js
mareki2p Question for i2pd: In file I2CP.cpp, in function I2CPSession::ExtractMapping, on line 583, in expression if (buf[offset] != '='), is there possible out-of-bounds read? The same applies to line 592.
mareki2p I have another question for i2p: In file DSAEngine.java, in function signIt, on line 433, in expresson `if (rbytes.length == 21)', could this ever happen? The same applies to lines 438, 452, 457. Because both r and s numbers are moded by dsaq, which is itself 20 bytes long number.
zzz it's a java biginteger thing
zzz anybody want to set up a sticktock.i2p? sticktock.com
dr|z3d you do :)
zzz not my expertise, and not in right country
dr|z3d looks like the copy button and associated code can also be removed from the b/w tester without issue.
dr|z3d or maybe not.
dr|z3d test runs fine, we just lose various datapoints like the server location and results.
zzz FYI there is a CLI for mlabrunner for rapid testing
dr|z3d I think I noticed that, thx.
dr|z3d More interested right now in what we can cull from the code without breaking anything in the UI.
dr|z3d Also trying to identify where we configure a 2 minute delay for router clients in the event the delay isn't configured in clients.config
dr|z3d mostly because a 2 minute delay before starting SAM leads to poor user experience.
dr|z3d we could probably leverage our Tor browser user agent for the b/w test so we don't reveal more info than necessary, and maybe spoof other variables that are being sent, like OS.
itsjustme Shizzle dizzle
dr|z3d damn that's some latency you're running there, ijm :)
dr|z3d all the middlebox code can go, zzz, tested, no issues.