Jaromil:
> any thoughts about this new systemd-made thing that freedesktop
> immediately "standardized" (whatever is their procedure for that,
> likely smoking cigars among old-boys or so)
> https://www.freedesktop.org/software/systemd/man/machine-id.html
...

. It doesn't say "why".
Why should I (we) provide a means to uniquely identify my systems ?
I don't need it. Is it for
- the nice though not useful concept, yes we reinvent things you know
- the ongoing trend for more surveillance from the authorities
- the ongoing trend for more surveillance from the advertising firms
- the ongoing trend for more surveillance for "copyright" reason

. It says "This ID uniquely identifies the host."
- on the lan I can just as well use hostname or the primary ip-number
- on the internet I use the fqdn or the ip-number
The uniqness part will fail without someone guarantying the uniqness.
Do we have an "IEEE handling out oui's" - for machine-id's ?
Even if there were some IEEE, pci- or usb-consortium handling out vendor
parts of the id's, there is no guaranty for uniqness since that
number is just what's the local admin fancies, or is it for
"tamperproof" hardware, something oppressive regimes would like.
And I have seen pci cards with the wrong vendor part: "the programmer
left and we don't have the code so we cannot change it" as the vendor
put it.

> AFAIK chromium started checking it and its absence on Devuan Beowulf
> is reported as an error, so we may have to work around this.

Isn't that a bug, the above man page says:
It should be considered "confidential", and must not be exposed in
untrusted environments, in particular on the network.
Why would a network centric program need this unless it wants to
expose it ?

> but first things first: do we want /etc/machine-id? and how?

In my view it falls in the completely unnessesary or the potentially
dangerous group.

I don't want it.

Regards,
/Karl Hammar