Re: Current behavior to set DSCP EF code point by default is harmful
Colin Watson <[email protected]> Sun, 12 Apr 2026 17:26:42 +0100
| Newsgroups | gmane.network.openssh.devel |
|---|---|
| Message-ID | <[email protected]> |
On Sat, Apr 11, 2026 at 11:58:07AM +0000, Job Snijders via openssh-unix-dev wrote: >On Sat, Apr 11, 2026 at 10:51:51AM +0200, Hendrik Visage wrote: >> Well... since Trixie/Excalibur (Debian 13 / Devuan 6) I've seen some >> SSH network "hangs" which might just be explained by these bits and >> the networks I cross. > >No, you have now in fact added a point to the contrary: these "hangs" >you report are entirely unrelated to any of the changes under >discussion, because Debian Trixie shipped with a debian-specific set of >patches applied to version of 10.0. I think Job knows this, but for completeness: I dropped the relevant patch post-trixie, given the changes in OpenSSH 10.1. That patch is one that we had been carrying for quite some time, since Debian 10 (buster), so I share Job's scepticism that behavioural changes between recent Debian stable releases can be chalked up to this. (That patch was an attempt to solve integration problems by reverting to the behaviour of OpenSSH < 7.8, but I'm willing to accept that it was a mistake at the time.) I'm not necessarily opposed to backporting the relevant changes to trixie and dropping the relevant Debian patch, but I make no pretence at being a network engineer, so it really needs to be an uncontroversial change or at least one I can have help in persuading Debian's stable release team to accept. I think the relevant commits are 65909fa114e7dd7511800db2b7bacb8774afe887, ec3465f59c651405e395092f3ad606f8992328d8 (maybe), and 289239046b2c4b0076c14394ae9703a879e78706, but I would be happy to take advice on that. This thread doesn't make it 100% obvious that such changes would be uncontroversial. I can't find any Debian bug reports indicating problems with QoS since we adopted the changes in OpenSSH 10.1, which is useful to know though not definitive (it took about eight months for reports about iptables and VMware problems with the changes in 7.8 to arise). -- Colin Watson (he/him) [[email protected]]