#1101950 Please consider a configuration-only non-default package systemd-resolved-mdns

Package:
systemd-resolved
Source:
systemd-resolved
Description:
systemd DNS resolver
Submitter:
Josh Triplett
Date:
2025-06-07 13:13:02 UTC
Severity:
normal
#1101950#3
Date:
2025-04-02 18:52:23 UTC
From:
To:
(Please note, this is a wishlist request, and low priority. It is not
meant to make additional work, and can wait until the current situation
calms down. Thank you for all your maintenance work.)

Since systemd-resolved -9 uses configuration to disable the mDNS
functionality by default (satisfying the requirement that the "default
installation" of systemd-resolved have mDNS disabled), would you
consider providing a *non-default* package systemd-resolved-mdns that
contains only a configuration snippet turning it back on (numbered to
come after the one in the base package turning it off)?

That way, users whose configurations do want to use mDNS with
systemd-resolved could depend on `systemd-resolved-mdns` and have it
Just Work.

#1101950#8
Date:
2025-04-04 05:46:56 UTC
From:
To:
Hi Josh,

That's an interesting idea. On the CTTE side, it was discussed whether
systemd-resolved could Conflicts: avahi-daemon and that was ruled out
rather early. I do not see how this would carry over to
systemd-resolved-mdns though. There, that conflict would really make
sense rather than breaking use cases.

Or it could go one step further still and we could have both
avahi-daemon and systemd-resolved-mdns declare Provides and Conflicts
for a virtual mdns-resolver package for others to depend on.

The CTTE questioned whether avahi-daemon would be the default resolver
in the long run, so this approach would also enable a way to eventually
transition the functionality in a smooth way.

I agree that this is all not urgent and we may defer it after trixie.
Yet, it is an interesting way to think about the problem. Thank you.
This is all my own views without a CTTE hat.

Helmut

#1101950#13
Date:
2025-04-04 23:41:34 UTC
From:
To:
Helmut Grohne wrote:

That was exactly my hope.

I don't think a virtual package will work well, because there's no standard interface for publishing things via mdns. And for resolving, the only *standard* interface available is NSS, so that would be something to handle via the corresponding libnss packages rather than the daemons.

Much appreciated. I'm glad it helps.

#1101950#18
Date:
2025-06-06 16:30:25 UTC
From:
To:
Would it be an option, within the CTTE decision, to move the
00-disable-mdns.conf file to avahi-daemon? That would still have the
same effect as long as both are installed, but would remove the need for
an extra enable-package (because then on systems w/o avahi, which are
the ones where using an enabling package would make sense, mdns would
already be active).

BR
chrysn

#1101950#23
Date:
2025-06-07 13:03:45 UTC
From:
To:
Hi chrysn (2025.06.06_16:30:25_+0000)

I'm afraid not. That was a different option on the CTTE ballot [0]
(option A) and it did not win.

[0]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1091864#136

If the avahi-daemon package maintainers had been amenable to this
option, it may well have won, but they rejected it. So selecting this
option would have required overruling these package maintainers.

Stefano