Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 23 Apr 2020 03:44:17 -0400
From:      Aryeh Friedman <aryeh.friedman@gmail.com>
To:        Ihor Antonov <ihor@antonovs.family>
Cc:        FreeBSD Mailing List <freebsd-questions@freebsd.org>
Subject:   Re: Wayland on FreeBSD
Message-ID:  <CAGBxaXn=x1mcQxrArE5v9-u9nm63BbTW+5t04XH_gpdhMTo4FQ@mail.gmail.com>
In-Reply-To: <5058973.kMyvyFPq5o@amos>
References:  <CAFYkXjmfyLZAi1HZe-RE3wLxa6GRNP6GkmtZG-4T2puRDOz0JA@mail.gmail.com> <CAGLDxTX5EeL3YDUJocdOM03sRzUDi3ed9cKuNH99DieZbrhGHg@mail.gmail.com> <5058973.kMyvyFPq5o@amos>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
On Mon, Apr 20, 2020 at 4:25 AM Ihor Antonov <ngor@antonovs.family> wrote:

>
> BUT FreeBSD folks are conservative bunch, some especially don't like
> changes
> that come from "linsucks". Others say that FreeBSD is a server OS. Same
> "server OS advocates" comfortably sit behind their MacBooks. There are
> ones
> that  make a living with it so any unnecessary change at the very least is
> ignored or at most is actively being not welcomed.
>

I want a beat up old pickup not a sports car to carry lumber in ;-)   (Is
FreeBSD sexy? no ... Do I care? no)

Also Linux (and yes I am resisting calling it by it's proper name you gave
above) has broken an amazing number of very simple things completely in
half to the point of making them unusable.   Let's take an example that
caused me to lose a $20k contract back in 2014.   Client wanted me to write
a "decent" installer for OpenStack, this relates to Linux see below, (I
don't know if the installer has improved since 2014 but back then it was
horrific to put it politely).   So I decided to start with DevStack (*NONE*
of the tutorials on how to do a hand configuration of OpenStack worked and
none of them {nor the OpenStack documentation}).  So once I set it up I
decided to reboot the host and no matter what I did after that DevStack
simply refused to work.    After a week of digging I ran across the
following passage in the manual:

"The worst disaster for a cloud is a power loss." (
https://docs.openstack.org/nova/latest/admin/node-down.html).

So I decided to dig deeper and found out the problem was iSCSI (a Linux
feature not OpenStack one) and its completely brain dead decision to allow
block level access to disks over the network and do so as if it was a local
disk as far the filesystem was concerned (with zero safeguards for the
connection being lost, which makes a corrupted disk/file an almost
certainty in the case of a power failure).   So means if Linux had decided
to do network storage the *RIGHT WAY* it have would of done something like
NFS (if not NFS it self) and the worst that could have happened  to
OpenStack from a power failure would have been to force the guests to run
fsck.    Linux making the choice it did of iSCSI over a NFS-like system
makes it impossible to run a stable cloud outside of a purpose built data
center that can guarantee no power loss for no more then 1/10 of a second.
(Doing so is not a good option for many organizations for security
reasons.)   Whereas even though PetiteCloud (see my signature) is nowhere
near complete or ready for third party use, it already kicks the ass off
OpenStack in that it can suffer a power loss without even batting an eye.
I have been using it in my home office as my private cloud system for
almost 5 years now and when I vacuum or there is a t-storm I lose power
frequently and have never had to do anything more then a simple fsck of a
guest drive despite having the file server and my desktop on physically
separate machines.

TL;dr; due to a royal F'up in basic storage system architecture Linux is
unusable outside of data centers for any sort of cloud computing whereas
FreeBSD doesn't even blink an eye in such environments.   So you tell me
which one is more modern and useful?


There are many more examples, these are just the fresh ones. Read arch
> mailing
> list, I am not making this up). A few progressive developers are
> outnumbered
> by fanatics of retro-computing. "Don't fix what is already broken" they
> say,
> "nobody uses it anyway" :P


A rush to be cool is usually a recipe for disaster unless you have no
serious work to do on you machine except to paint racing strips on it (and
the OS) and install flashing mutlicolored LED strips along the mother board
because it "looks cool".

I just gave a prime example above where the desire to be on the cutting
edge without really understanding what had already been done before and
proven to work completely crippled a product.   I forgot to add that the
mistake I mentioned above was so awful that it led NASA's inspector general
to conclude that OpenStack was a $35M complete waste of tax payer's
money!   (https://oig.nasa.gov/audits/reports/FY13/IG-13-021.pdf)   It
should be noted like typical Linux developers OpenStack never paused for
one second to go back and fix a glaring and obvious design flaw.   Instead
they just piled more and more ill-conceived and even worse-implemented crap
on top of the flaw making any future failures even worse (and harder to
fix).

The FreeBSD core team would have *NEVER* tolerated such complete disregard
for even the most basic system/software quality assurance!



>
>
> I am afraid that on FreeBSD Wayland will remain a third-class citizen.
> Look at sway for example: it needs Linux Kernel event API (evdev, luckily
> on
> by default in 12),  epoll-shim and eudev-shim, and then some magic with
> barely
> documented kern.evdev.rcpt_mask=12 in sysctl.conf to make it work.
>
> Why? Because FreeBSD folks are generally not interested in contributing
> proper
> kqueue and devd support to "linusucks" projects. As well as maintaining
> X11.
> Linux developers are not going to write that either. The amount of
> linuxisms
> and shims is only going to grow, the desktop team is fighting an uphill
> battle
> here, there is no proper infrastructure for them to give users good
> Wayland
> support. All they can offer is shims, hacks and workarounds.
>
> And so Wayland will slowly but surely replace X in Linux world. But even
> in 10
> years FreeBSD hackers will keep using rotting X11, will keep stating that
> they
> had jails first, ZFS is cool, GPL is evil and Netflix uses FreeBSD on
> their
> servers.
>
> FreeBSD is not a general purpose OS. It is a server OS, It is retro OS. It
> is
> OS that fills the niche where closing sources is needed for commercial
> success.
> An OS you want to put somewhere and never ever touch it again. It is not a
> desktop OS, quoting someone from this very mailing list: "It is power to
> serve, not power to desktop!"
>

I challenge you to show a single item in the base system that does not have
the source available in a better documented and easier to find place then
Linux does (any system that can only be built easily from binaries is not
very open!).

BTW as far as desktop use goes it does everything I need right out of the
box with no special configuration needed after installing the XFCE ports
(plus any browser and office apps I use).


-- 
Aryeh M. Friedman, Lead Developer, http://www.PetiteCloud.org



Want to link to this message? Use this URL: <http://docs.FreeBSD.org/cgi/mid.cgi?CAGBxaXn=x1mcQxrArE5v9-u9nm63BbTW+5t04XH_gpdhMTo4FQ>