:: Re: [DNG] What is a user-init?
Page principale
Supprimer ce message
Répondre à ce message
Auteur: Alexandros Prekates
Date:  
À: dng
Sujet: Re: [DNG] What is a user-init?
On Sun, 10 Dec 2023 02:22:42 -0500
Steve Litt <slitt@???> wrote:

> Hi all,
>
> I never heard the phrase "user-init" before the emacs daemon thread.
>
> 1) What is the definition of a user-init?
>
> 2) What requirements must a user-init fulfill to be acceptable?


If we accept the posibillity that a user would want to start
daemon like programs (so that means a different kind of program that we
usually relate to a user that we usually think would do word processing,
video editing , gaming ,browsing and generally information processing
tasks that involve close-frequent involvment of the user in the
processing loop and not automation ) then i think a need arise for
a supervisor of those user-daemons and of course a need to initiate
those user daemons (handling dependencies, initial config etc) .
So that scenario can map to a machine on top of which run
multiple virtual private servers(vps). Each vps has each own init that
set up a vps to a certain service . So each vps's admin effectively has
his own root-init but for the perspective of the vps-es supervisor admin
she could be forgiven to see all those vps system admins as user of
his bare metal server and their init as user inits.
So in that context i think that it's a matter of perspective and where
you put emphasis (or focus) that enhances a certain semantic
interpretation of the word init as referering to either the initiator
of the basic non-kernel processes and daemons that create a base for
a userland or as refering to the initiator of a user-init that wants
to set up a mail-server, a web server or etc.

So i think , having as stimulus the emacs-server that felt like
an outcast of both the init-space and the user-space (as i usually
was thinking about it) and the dbus architecture that in a way
allows for userland processes to act as daemonts-servers to other
userland consumers i put forward a general issues as to whether
we are in era of 'personal computing' that a user could have
her(his) own user-init that would initiate her(his) daemons
and so in a pc could run simultaneously more that one user-lands
each with her own daemons and thus different user-inits and supervisors
and different dbuses.

We could think that scenario as a ligthweigth version of the vps
scenario i said before. Light in the sense that i feel that a certain
OS could have restrictions in how flexible it can be to support
in paraller unrelated daemonts workloads . Its one thing
to have two desktop event daemons or emacs-like app-daemons working in
the same machine for different users and another thing to as a machine
to be at the same time a game server and a mail server.

So is there a need for ligthweight user-inits and user-daemons
supervisors ? I think that that development would offer semanticaly
more sound systems less prone to schizophrenic semantics. (how else
should i call the case of having init setup a user service without
a user login..) The root should not handle user-services but only
system-wide services that support the most generic userland.
If a user wants to transform his vps in a certain server s(h)e should
do that from his user-init. There should be a seperation between
init-supervisor that maintaints and supports a general userand and
the health of the system itself and the user-init-supervisor that
could go down without endagering the system.
So in a sense the way things are done now could be seen as promoting
an integration in how we see a 'system' and not modularization.

Alexandros.