My apologies, this was borne of frustration at a number of tensions
breaking at once. That said, on calmer reflection, I do have a couple of
suggestions for this document to perhaps mitigate the situation somewhat:
- Reconsider whether the Tip about networkd is necessary. While the text
does say "may", its prominence at the top of the document draws
attention to it, giving it a sort of pride-of-place.
- Replace "modern" with a different adjective. It's been my observation
that people, especially in tech, see this word and associate it with
being somehow better. They might also see it and think that this is the
direction things are headed. I think "alternate" or "alternative" could
work here to convey the available options without evoking a value judgement.
Hope this is of some help.
On 9/11/25 9:51 AM, Osamu Aoki wrote:
control: retitele -1 provide fair warning to Bug1112535-like situation
control: severity -1 wishlist
thanks
Hi,
Thanks for the reminder.
I see 4 references to systemd-networkd feature.
Please note I use "MAY". So it is presented as optional feature.
Disclaimer: I am only following systemd-networkd(8) documentation to write this
section. I don't use this feature.
As I read mentioned bug report #1112535, Luca doesn't mention his rationale for
his "EXPERIMENTAL" judgment. I don't think this bug title "maintainer insists
systemd-networkd not to be used" is fair characterization for discussion with
Luca.
Since upstream systemd doesn't say "EXPERIMENTAL" for the systemd-networkd
feature, it is in reasonable shape by itself, I suppose.
I understand Luca may call use of systemd-networkd as "EXPERIMENTAL" from system
integration perspective because it changes such a fundamental system component
which is setup by debian-installer from "Debian system" perspective. Unlike
bash-to-dash or usr/bin-merge, this is not on the table for migration and not
tested for upgrade.
I think it is wrong to press package maintainer and release manager for bug
management.
I can think of many similar alternative components with similar essential
features (use of dracut-core instead of initramfs-tools). Upgrade robustness
tend to become fragile. Also, it is impractical for package maintainer to
guarantee upgrade stability. Fair warning somewhere for making wild system
change from d-i installed state.
Let me think how to address.
Osamu
On Tue, 2025-09-09 at 12:12 -0400, Kevin Otte wrote:
Source: debian-reference
Version: 2.127
Severity: minor
Dear Maintainer,
Section 5.3 of the manual clearly outlines the usage of systemd-networkd for
non-GUI network configuration under Debian.
Its inclusion here indicates this is a supported mechanism.
However, in #1112535 the systemd package maintainer, Luca Boccassi, insists
that systemd-networkd is an optional component, at one point even going so far
as to call it experimental.
If Luca is correct and this is the case, the manual should be updated to
reflect this fact.
If the manual is correct, as I believe it to be, Luca should be informed of
this and encouraged to issue a correction.
Many of us in the mentioned bug report have attempted to persuade him of this
to no avail. Perhaps you will have better luck.
