On Mon, Sep 25, 2023 at 2:34 AM Dave Close <dave@compata.com> wrote:

I wrote:

> I detest a graphical login and insist on running in multi-user mode
> (old runlevel 3). To start an X11 session after login, I can use the
> startx command. But I haven't found an equivalent command to start a
> Wayland session.

Have you tried <https://wiki.archlinux.org/title/GNOME#Wayland_sessions>:

if [[ -z $DISPLAY && $(tty) == /dev/tty1 && $XDG_SESSION_TYPE == tty ]]; then
  MOZ_ENABLE_WAYLAND=1 QT_QPA_PLATFORM=wayland XDG_SESSION_TYPE=wayland exec dbus-run-session gnome-session
fi

The article has note that "The factual accuracy of this article or section is disputed."

Tim via users wrote:

>I don't really see the advantage in not starting a desktop session,
>just to log in, then starting one afterwards.  There must be something
>I'm missing.

[...]
The principal advantage is that I get to see what the system is doing
during boot and login. I strongly dislike any interface that tries
to make things "simple" by hiding what it does. I don't trust things
that are hidden from view.

I always remove the "rhbg quiet",  journactl is a big improvement over

trying to figure out which log file is relevant to a problem.

 

And I've probably used computers longer than most of us, having started
with an IBM 7094 and an 029 keypunch.

As an undergraduate in 1968 I had access to <https://en.wikipedia.org/wiki/Dartmouth_Time_Sharing_System>

using a teletype terminal and saving programs on paper tape.  It wasn't until grad school in 1972 that I met

a keypunch.

 

There are certainly good use
cases for a graphic application, especially for things I don't do
often and don't remember the tricks. But every graphic system has
its own ways whereas the command line uses only things on a standard
keyboard.[...]

 

These days, many linux users never use the command-line until something breaks

and the smartphone images they post cut off key parts of error messages. 

Linux is community driven.  The community now includes kids whose first

experience was a smartphone.  I used to run "practical" sessions for a

workshop where a linux application was the only implementation of the theory

being taught. We found is necessary to spend the first two afternoons introducing

the command-line so later sessions weren't dominated by problems with linux.

The community benefits when experienced users regularly use the most

common configuration (e.g., Fedora Workstation with Wayland) so they can

identify issues and pitfalls as well as encouraging new users to learn POSIX

shell commands in a terminal.

I agree and disagree with your position, but for different reasoning.

When possible, I like to use a graphical environment.  Its simpler and faster for most everyday things.

However, I strongly dislike the make-it-look-like Windows thinking that goes on at RedHat and by association, Fedora Workstation.  For example, the "rhgb quiet" boot defaults make it hard to actually see what is going wrong when you have a hardware issue, a misconfigured app or service, or a bad update.  When you have something that only fails once in a while like a race condition issue, the boot messages should be in your face, not hidden behind a stupid spinner screen that then locks up.  The Gnome software app needs a complete rewrite, and I far prefer the faster and more accurate feedback from the CLI dnf command.  Maybe the GUI is fine for installing common apps, but for example, there is no entry for the thunderbird-wayland package that is required to run Thunderbird in F38.  If you have a Gnome desktop or app fault, it is extraordinarily difficult to debug, most often requiring a special debug version to be installed to even give you basic logs.  If you need to see things like disk i/o performance, the system monitor app does not give you that information, and you have to go to the CLI with applications like top and iostat.

I'm also perplexed that Fedora server uses weird things like a different filesystem than Workstation.  That's just wrong.  Btrfs is as good as zfs, and a considerable improvement of xfs and ext4.  I've had nightmares trying to patch an xfs filesystem because after 50 years of development, it is still not stable.  Btrfs is at least as fast at both read and write, and does a far better job handling bitrot on modern disk sizes.  Xfs cannot do that, and probably never will.  The toolset is also richer, making things like backup easier, faster and more reliable.  Using xfs as the default fs is an obvious deferment to its RHEL cousin where questionable reasoning ends up in bad decisions, and holding back an excellent filesystem like the superior btrfs fs for unclear reasons.  This very questionable decision is also the underlying reason why an ext2 partition and grub have to live on, instead of booting from a single btrfs partition and making disk management so much easier.  Complexity in a server implementation is bad IMHO.

While I am a strong supporter of Fedora, I appreciate the more correct thinking behind the information feedback given in Debian and some of its derivatives like Ubuntu.  When something goes wrong on those distros, it is far easier to recognize and correct than in Fedora or RHEL.  IMHO, Fedora needs to adopt much of the Debian thinking and stop making a Windows clone.  People who move to Linux are not interested in making it work like Windows, but instead want something better.  Fulfil that promise instead of hamstringing them.  Bad decision making is one of the reasons why IBM (who own RedHat and many of the Fedora devs work there) is mostly irrelevent today.