absurd and long /usr discussion on gentoo-dev

ulenrich · Guru Joined: 10 Oct 2010 Posts: 468

There is an absurd discussion on the mailing list about the

upstream move to /usr (sake of distribution_single_point_ro)

For the whole lot of people there is no change but a shorter PATH environment.
Only people who want
- a seperate partition of /usr
- not to use an initramfs
are affected. These people want to keep the old /-root directories of
/lib /bin /sbin (which will be removed by upstream).

But there is a very,very simple solution for these people: Why not a new USE varible
seperateusrbutnoinitrd

Which could be interpreted by portage to just copy a bunch of libs and bins into old /-root direcitories and to keep the long PATH environment. This way we all keep up with upstream and these few people do have their free of any initrd boot possibility?
_________________
fun2gen2

cach0rr0 · Posted: Mon Apr 23, 2012 4:00 am Post subject:

because
a)it breaks shit
b)upstream has the attention span of a methamphetamine-crazed chihuahua

everything "upstream" is doing seems to be almost exclusively for the benefit of GnomeOS, with the mind that everyone else is simply going to fall in line.
If i wanted to use GnomeOS, i would not use gentoo.

There will be a path provided for people to go about things either way; however it grows very tiresome tracking a new "standard" introduced by the Good Idea Fairy that requires a complete gutting of everything you do, only to find out 6 months later the Good Idea Fairy has changed its mind and the "standard" is now to be something entirely different.

Some discussion on the topic here: http://www.gentoo.org/proj/en/council/meeting-logs/20120403.txt
whoever Chainsaw is, his take on the matter is buy-that-man-a-beer-worthy
_________________
Lost configuring your system?
dump lspci -n here | see Pappy's guide | Link Stash

ulenrich · Guru Joined: 10 Oct 2010 Posts: 468

Same procedure as every 5 years?
Next fork of gentoo is pending: "exzerbo"

This was said and keeps valid as answer of the council decision:
"Unless somebody is actually willing to maintain a robust alternative I
don't really see that as a real option. If upstream moves in one
direction, and nobody is willing to maintain things in a different
state, then you just end up with a system package that nobody wants to
use, and something in an overlay that everybody uses instead that is
beyond these debates. You can't effectively mandate that people
maintain something in a volunteer organization, unless the effort
involved is very minor."

My point was:
The other way round is easier to maintain!
And error free ...
_________________
fun2gen2

Dont Panic · Posted: Mon Apr 23, 2012 4:44 pm Post subject:

I think the council would love to have a roadmap for something comparable to a 'seperateusrbutnoinitrd' USE flag.

The problem they keep running up against is that will involve forking some important packages (udev being the first one). As time goes on, more and more packages will require patching/forking since upstream will just assume /usr is mounted.

Feel free to elaborate on your 'seperateusrbutnoinitrd' USE flag solution. Many people would salute you if you can find a way to make that work without a high maintenance burden.

depontius · Advocate Joined: 05 May 2004 Posts: 2156

There are annoying forces at work here, and this debate is one symptom. Basically it looks to me as if people who grew up on Windows and are comfortable with everything that implies, including the Registry, have moved to Linux and are trying to make it more "comfortable". Some of the legacy Linux decisions and design points look uncomfortable to them, and they're trying to change that. Or at least that's what this all feels like, to me.

I won't deny that Linux can learn some things from Windows, but I would also assert that Linux has learned some lessons that Windows hasn't, and some of those "uncomfortable legacies" are in there for problems Windows users never thought of hitting, yet. The goal of "just works" (TM) is great, but the real problem has always been that it doesn't always, and usually by the time it doesn't always, what's there stymies you horribly and seems to actively work to prevent you from fixing your system.

Yet at the same time, these changes are coming from RedHat, and it can be hard to fight city hall.

So I focus on 2 things:
1 - Documentation - I really hate those "magic strings" from freedesktop.org - the kind where the answer makes perfect sense after the fact, but you have no idea how to get there from ground-zero. There is career-path documentation available at freedesktop.org, but nothing for someone whose system doesn't "just work" (TM) and is simply trying to get it there. We need some decent mid-level documentation.
2 - Debug and non-GUI access - One example of this would be DBus, as others have mentioned being able to see and especially filter the bazillions of DBus messages wandering back and forth. (Perhaps one of the most important aspects would be to see the DBus messages that nobody is listening to, and figure out why not.) Another example would be the binary turn that systemd is taking with its "journal". We need a bunch of text-mode (and GUI) "fuzz" programs around this new stuff.

Once again, the goal is to sit at a system that doesn't boot properly, figure out why not, and FIX IT!

In the long run, I don't think it will be productive to dig in one's heels, refuse to accept DBus, PAM, systemd, *kit, etc. Of course you can roll your own alternatives, many have. But unless you also bring such workarounds to full "OSS project" status, grow a community, and become a realistic fork, you're tilting at windmills and will eventually be alone in your own sandbox. The alternative is to take these "newfangled gadgets" and surround them with the utilities we know they should have.
_________________
.sigs waste space and bandwidth

cach0rr0 · Posted: Wed Apr 25, 2012 5:25 pm Post subject:

cach0rr0 · Posted: Wed Apr 25, 2012 5:44 pm Post subject:

Ant P. · Veteran Joined: 18 Apr 2009 Posts: 1923 Location: UK

NeddySeagoon · Posted: Wed Apr 25, 2012 6:12 pm Post subject:

cach0rr0++
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

MarekSieradzki · n00b Joined: 05 Oct 2010 Posts: 17

cach0rr0 · Posted: Wed Apr 25, 2012 6:42 pm Post subject:

MarekSieradzki · n00b Joined: 05 Oct 2010 Posts: 17

MarekSieradzki · n00b Joined: 05 Oct 2010 Posts: 17

cach0rr0 · Posted: Wed Apr 25, 2012 6:58 pm Post subject:

jemi · n00b Joined: 06 Oct 2010 Posts: 21

It's pretty silly that they wouldn't want a flexible filesystem standard. Do ubuntu users really care if they won't be familiar with my Gentoo systems filesystem layout? Mostly no. Do I want my gentoo system designed so ubuntu users are more comfortable on my system? Absolutely not! The fact is, windows has one way of doing things, and linux has a whole bunch, and to assume that this "compatibility" issue is important when it doesn't even meet reasonable definitions of the word "compatibility" is really silly. Is udev guaranteed to be a part of every posix system? No. Is gnome? No. If there is a way to avoid adopting this standard in Gentoo by forking udev or something like that, I'll be all for it. I say let's resist this until upstream gets their head screwed on straight.

NeddySeagoon · Posted: Wed Apr 25, 2012 9:02 pm Post subject:

jemi,

Lots of people agree that the direction being taken by GnomeOS is a VeryBadThing but they are not writing code - GnomeOS is and thats they way it will contine unitl
a) nobody can remember any different
b) the people who agree about the VeryBadThing write some altenative code of their own.
c) GnomeOS changes direction.

Just now, only a) looks likely.
b) is unlikely becuse its not just udev - its lots of other applications too.

I'll live with an initrd to have a separate /usr and /var, I mostly use root on lvm on raid anyway, so I already have an initrd.
What I won't live with is the initrd being produced automatically by some black magic I neither understand nor can fix when (not if) it breaks.
I won't have any GUIs on my servers either.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

avx · Advocate Joined: 21 Jun 2004 Posts: 2064

Simple, kick out GNOME (and to a degree, KDE) of portage, fix the rest - if there's one. People desperate for the stuff can make an overlay or switch to GnomeOS.

Funny thing is, I switched to Linux, because it was easy to be free, not that it's made complicated by a Linux company, I might as well switch back to the originals (MS/Apple).

To quote Poettering, "it's open source, send patches!": that's (mostly) out of my league and those who tried, were denied.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>.

saellaven · Tux's lil' helper Joined: 23 Jul 2006 Posts: 113

salahx · Guru Joined: 12 Mar 2005 Posts: 349

This whole is less a big deal than its being made up to be. There are 3 kinds of configuration:

Those who have both / and /usr on the same partition - no initramfs needed, unless / is on something fancy, in which case they are using one anyway. No change in configuration.
Those with separate partitions - / is only something fancy - initramfs required, they are already using one. No change to configuration although the initramfs will have to be modified to mount /usr)
Those with separate partitions - / is vanilla - initramfs required, wasn't before.Configuration change required If /usr is vanilla, a minimal initramfs is needed; if /usr is fancy, they'll need the full strength initramfs.

("fancy" means on software RAID, LVM, NFS, and so forth)

The last group will be slightly inconvenienced as a result of this change

The Doctor · l33t Joined: 27 Jul 2010 Posts: 948

Havin_it · Posted: Thu Apr 26, 2012 11:39 am Post subject:

avx · Advocate Joined: 21 Jun 2004 Posts: 2064

Havin_it · Posted: Thu Apr 26, 2012 7:40 pm Post subject:

avx · Advocate Joined: 21 Jun 2004 Posts: 2064

Still, KDE is somewhat helping Gnome in this situation I call (at the very least) problematic. F.e. dbus, by also picking it up, I guess ~85%+ of DE developers and their applications are now behind this and thus forcing it on the people developing not so mainstream applications and windowmanagers, because it gets deeply built into core libs of kde and gnome, which other projects also use.

Years ago, in the times when 'one app for one problem' was still the way to go, it's been easy to mix and match, since all these apps communicated in easily useable and predictable ways. That times are practically over, now that many projects try themselves on the monolithic approach. There wouldn't be much wrong with it, if these apps wouldn't be core applications/libs pretty much anybody is forced to use to have a working system.

Such behaviour is good on the sides of MS/Apple, which make a living by setting standards how to use and design things fitting in there respective philosophies, but for Linux, where people always said diversity is good, this is really really bad.
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>.

Ant P. · Veteran Joined: 18 Apr 2009 Posts: 1923 Location: UK