Re: [gentoo-user] X11 without udev/eudev

[Canek Peláez Valdés]

On Mon, Aug 23, 2021 at 11:07 AM <karl@××××××××.se> wrote:
[...]

> It is inconvenient that thoose two goes away. Regarding udev, it has never
> supported serial mice, so it doesn't help me.
>

What are you talking about? Udev doesn't "support" any hardware; as the
manual page states[1], it: "supplies the system software with device
events, manages permissions of device nodes and may create additional
symlinks in the /dev/ directory, or renames network interfaces".

If the kernel supports the hardware, udev will supply the corresponding
event, so udev will generate the corresponding event for serial mice also
(probably close to boot time).

[...]

> Poeple write whatever software they want to or are paid to do. It is my
> call if I want to use that software or not.
>

Well, yeah; but if you want to use software X, and X depends on Y, then you
either use software Y or don't use software X. The dependency chain can and
will grow depending on several factors, and it's the situation here if I'm
not mistaken: you want to keep using keyboard and mice, those depend on
libinput, and libinput depends on udev.

There are alternatives, of course; some members have posted several working
ones; but the fact is that either you accept the dependencies dictated by
the software you want to use, or you need an alternative that *someone* has
to write/debug/maintain, etc. You said: "[u]dev should be an optional
daemon, utilized when the local administrator decides to do so," and that
is simply not true: if a software you use decides to depend on udev, then
you either use udev, or need to change to a different software.


> As I wrote before, udev does not handle serial mice, so udev does not
> solve anything for me nor does it help me in any way to run my systems.
>

You are saying it wrong, you mean to say:  "to handle serial mice, I don't
need udev". And that is 100% completely true; you can keep a static /dev,
don't use udev and create the device nodes by hand. But serial mice
work great under udev also.

The thing is, the *GREAT* majority of users need support for dynamic
hardware, so the developers of most of the software in Linux solve the
general problem: dynamic hardware (which includes static hardware also). So
they do that, and you complain about daemons that are needed for other
people other than you.

But you didn't write the software that depends on udev; someone else
(worrying about the general problem) did. So either you write your own
version, or find someone else who does (again, check the alternatives).


> Udev is just something pushed on me for no gain except possible to satisfy
> some dependancy touted to be beneficial.


For no gain *TO YOU*. You want software that someone else wrote to cater to
*ONLY* your needs. That's not how it works; *MOST* developers will cater to
the needs of the *MAJORITY* of users, and that includes the dynamic
hardware scenario (which has been the case for all of the XXI century, by
the way).


> So in this very specific case it can be considered "bloat" if you wish to
> use that kind of words.
>
> My guess is that it is more useful on laptop than on a desktop box or an
> industrial computer.
>

Of course not; desktop and servers ("industrial computers") need dynamic
hardware (hot swappable storage, anyone?, disconnect the webcam after the
Zoom meeting?) Just because you in particular don't find it useful doesn't
mean it isn't: the truth is, dynamic hardware users surpass static hardware
users, and it has been the case for several decades now.


> As a side note, from what I understand, udev today is mostly
> about usb-devices because that is where the dynamic hardware comes
> from today (at least when we are not talking about hotplugging cpus, memory
> cards, io-cards and such (but that is more of a enterprise problem than a
> small system problem.
>

It's not only USB; it's Bluetooth, Thunderbolt, eSATA and eSATAp, etc. The
computing world is dominated by dynamic hardware; it has been so for at
least the last two decades.


> Serial ports are darn easy to implement in hardware and softwere.
>

Yes: but the software that *uses* mice doesn't care *ONLY* about serial
port mice. They care about USB mice (which is the majority) and Bluetooth
mice (which has the second place). Right now, serial mice have to be a
distant third place, if at all.

So the developers of software that deals with mice don't need to worry
*ONLY* about your case; they need to worry about the general case. Which
includes USB and Bluetooth (and whatever they invent in the future).

E.g. if I have a program connecting to a device using a serial
> and it is disconnected, I can just reconnect it and nothing
> special happens, noting to be done in software except logging.
>  The same device via usb, the dis-/reconnect will close the
> port and make it vanish forcing med to find out find out where the
> new /dev file is and reopen and reinitialize it.
>

That is what udev+libinput solves, *exactly*. You answered yourself why it
is needed *IN GENERAL* (not in your particular case).

What you say about complexity etc., is true; but most developers care about
servicing the *majority* of users (in this case, USB and Bluetooth users
included), so if you want to use their software, that means accepting udev.
If you don't like it, you need to stop using their software and either
write your own or find someone else who does it (again, check the
alternatives).

In any case, complaining is useless: most sane developers always want to
cover the majority of users. That's why udev is mandatory in most major and
medium distros; in Gentoo it is used by default (BTW, they are preparing
the deprecation of eudev in Gentoo[2]).

Regards.

[1] https://man7.org/linux/man-pages/man7/udev.7.html
[2]
https://archives.gentoo.org/gentoo-dev/message/dff4bf35636efef95f6d7926823b4e8d
-- 
Dr. Canek Peláez Valdés
Profesor de Carrera Asociado C
Departamento de Matemáticas
Facultad de Ciencias
Universidad Nacional Autónoma de México