Current path of Arch scaring me

kleptophobiac · 2006-02-25 00:25:52

I switched to Arch about two years ago because it was light and fast and clean. Since then, there have been a number of major developments. In order:

- Migration to udev. I approve of this one. It didn't really change jack from my perspective, seemed to effectively deal with my hardware, and wasn't much of a headache when it came down the pipe.

- Migration to initrd. I can't stand this one. I liked my arch kernels with their post boot loaded modules and simple boot images. It was really easy to turn them into PXE boot or flash boot kernels. Now I have ten bazillion modules all the time. Arch linux was always an additive process for me. I start with just about nothing and add what I want. The new kernels have been just the opposite... you start with six tonnes of BS and have to carefully carve away what you don't think you need and hope you don't break anything. It honestly feels like buying a computer off the shelf and having to purge it of crap. I know it's not tremendously difficult to do, but I don't think it's in the spirit of Arch

- Migration to xorg 7. I don't care so much about this one, and I understand that modularization of everything X is not the Arch dev's choice... but I don't like it. It seems to violate KISS. I'm just glad I don't use GUI's on more than two of my arch boxes.

Summary: udev good, xorg7 blah, initrd horrible

I guess I'm just asking peoples' opinions on these changes and how they feel they keep in line with the Arch way of doing things. Thoughts?

ozar · 2006-02-25 00:35:21

Yeah, I'm okay with udev and once the majority of the remaining bugs are worked out of xorg7, I'll be okay with it, too. However, I've never liked the path down initrd lane and don't expect that my feelings will change.

iBertus · 2006-02-25 00:41:35

I'm no longer using any stock Arch kernels because of initrd. I'm not tech savvy enough to know why it's supposed to be better.

Gullible Jones · 2006-02-25 02:58:39

You can get rid of the extraneous modules in your initrd. Specify the filesystem modules in mkinitrd.conf, disable SCSI and SATA if you don't use them, and set AUTODETECT to "1", then remake the initrd... voila, unnecessary modules are gone.

Regarding udev, the fact that it was shoved into the mainstream before everything was in working order really annoyed me, but now it seems fairly complete, so that's water under the bridge.

As for Xorg 7, I actually love it! Okay, it would be nice if the protocol stuff was all in one package, but there a probably very good reasons that it isn't. And I think it's great being able to skip the hideous 100dpi and 75dpi bitmap fonts...

codemac · 2006-02-25 03:22:52

kleptophobiac, excuse me if I laugh when your "Arch way of things" is simply a lack of understanding.

Cerebral · 2006-02-25 03:28:03

Re: Initrd - this doesn't bloat up anything. Before initrd, all these 'extra' modules you complain about having were compiled directly into the kernel -- you couldn't see them, but they were there, and you couldn't do anything about it. Now you can do something about it. Sure, the output of lsmod is a bit longer, but in essence you've got a lighter kernel running because you can now prevent these modules from loading at all.

Snarkout · 2006-02-25 03:36:05

I like udev, and am very glad arch uses it - I wouldn't use arch if it didn't most likely.

I like initrd and much prefer it to the "old" way.

I dunno about xorg7 - I haven't inflicted the pain on myself yet. Maybe this weekend.

cactus · 2006-02-25 04:21:42

kleptophobiac, I agree with many of your points.
Be aware, however, that many *users* in the community will percieve the fact that you dont like something, as some sort of deficiency on your part.

Be prepared for:
1) "omg. you newb. you don't know what you are doing"
2) "Do X instead of Y"
3) "Use another distro"

blah blah.

Again, I agree with you. I would suggest posting to the mailing list, and try to present your argument to the devs, in a presentable and well informed, civil manner. That will get you much better traction than posting in the forums..where "group think" seems to reign more supremely. (Though in honesty, the devs have heard it plenty from people like me, and seem to be set in their decision).

Good luck

Mikos · 2006-02-25 04:39:44

I am really happy that Arch uses X.Org 7.0! Same for udev and intelligent use of initrd (which is LESS bloat than before initrd, because all modules were compiled in kernel).

Dusty · 2006-02-25 05:51:59

I don't much linke initrd either, but it doesn't bother me enough to complain, especially since I know phrakture has been looking into better options.

Dusty

neotuli · 2006-02-25 06:45:53

Dusty wrote:

I don't much linke initrd either, but it doesn't bother me enough to complain, especially since I know phrakture has been looking into better options.
Dusty

yes he has. Ultimately look for initramfs to come and blow initrd out of the water. It's very nifty stuff.

dtw · 2006-02-25 09:38:29

Surely that is largely of the back of iphitus's work?

My only timy grumble would be how the direction of some of the major packages in the Arch repos seems to be decided on the current whim of a single developer. I assume that, whether Arch is for the devs or for the people, it wants to be "better" and I'm not sure if a single person always has the best POV to deliver that.

I'm not suggesting that Arch should be or would even work democratically, I just wonder if there is enough opopportunity for criticism of new ideas before they happen. Not all new ideas are good ideas. I can happily take into account the simple facts that change is always resisted, permenantly in some cases, and that sometimes it needs someone to just stand up and say "We're doing it this way," in order to get it done. However, while it is inevitable that a certain amount of pride will get mixed into the process, it concerns me when it seems to cloud up the process. If one person takes responsibility for a big change then they should also be happy to take the flak when that change is less than smooth. Devs are not infallible and there are some among us that have better ideas than the devs.

I hope that the expansion of the dev team will provide a foil for this apparently single-, almost bloody-minded approach to change and that the development of Arch continues in a series of large and firm strides rather than what has, in the last year, seemed a somewhat blind-folded stagger with numerous bumps.

I apologise if I seem vague. I am trying to be subtle yet constructive. This is my first foray into "Arch politics" for sometime and I hope it doesn't prove too painful for me!

vacant · 2006-02-25 10:27:18

I'd rate xorg7 as light fast and clean. Using xorg's radeon driver, FPS on my 9200SE went up from 950 to 1100 in glxgears. There was only one extra download for xf86-video-ati.

anykey · 2006-02-25 13:01:32

I am inclined to agree to that initrd rant; but then, I didnt use any of the stock kernels during the last two years (there were other things why I didnt like the kernel distribution in here).
To give you some advice: just pass on the stock kernels, and build your own; people who don't want to don't need to, but I want my own kernels.

One of Arch's biggest assets is -- certainly -- its adapatbility (of which abs is a great tool, along with the really good modular concept) to build yourself what the distro won't give you (no, you do not need to have every kernel upgrade, that's what changelogs might tell you).

As to xorg7, I think, this change was somewhat necessary. I am perfectly aware the transition wasnt too smooth. But then again -- I think, this was a one-time-glitch.

What I cannot stand is that attitude of some people here (which shall remain unnamed) to whom justified criticism equals blasphemy. Come on, folks, we mostly just want to use and enjoy it, don't we?

kleptophobiac · 2006-02-25 13:17:35

Well, as I said in my original post, xorg7 is in my opinion a "blah" change right now. I'd like to give it time to mature a bit. It's such a monumental change that the problems people are experiencing are inevitable. Also, I made a new install of Arch on one of my X boxen post the xorg7 update, and things went very smoothly. I wasn't happy to have to manually come up with a configuration file for X (the automatic generator disappeared! I don't know if it's back yet, but that was a little bit of a pain at the time). I like the modularity, I don't know if I like how fast it made it through testing.

I don't really know what was wrong with the original udev transition. I followed the instructions posted to the dev blog and had no issues (except on one machine where it was completely my fault). Udev rocks!

I'd like to see initramfs come out or a "regression" to the old way of handling kernels. The only things that really must be built into the kernel are IDE/SCSI drivers and filesystem drivers. Beyond that, everything can be modular. With the old scheme it was also trivially simple to edit the kernel .config and rebuild the package and have a drop in replacement. Things are a smidge messier with initrd. Also, I netboot computers. Initrd is more of a pain to netboot than a semi-monolithic kernel.

As for laughing about my understanding of Arch. Let me restate what I think "Arch is all about".

Arch is about simplicity. It is there not to make things hard or easy for the user, but to make it logical and simple. Arch has always been an additive process. You start with a naked base install and work your way up. You start with almost zero configuration and work your way up. My big concern is that with the introduction of initrd Arch is swinging towards being a subtractive process. You have a ton of modules and a bunch of extra hoopla to do because of initrd which you have to strip down to be efficient rather than build up to be useful. I think my analogy to an off the shelf computer holds true. When you buy an off the shelf computer you have seven layers of garbage you have to uninstall to have a clean and functional machine. It's tedious and has inconsistent results. When it comes to software, the additive process is much simpler than the subtractive process - and IMO, much less error prone. It's a question of philosophy rather than a question of initrd in particular.

pikass · 2006-02-25 13:19:23

Initrd doesn't really make things better it just makes some things possible (for ex. lvm raid1 for /etc or crypt keys on usb, in most cases server stuff). Maybe an optional Arch initrd Kernel would be more pleasant for many users.

From my point of view migration of udev and Xorg7 is out of the question. This had to be done sometime. It was only up to the devs to decide when. If they wait peoples would start to complain that Distros like Ubuntu are far beyond Arch (or even Debian ). So basically as a dev you can do what you want, only thing that is for sure, peoples are complaining.

codemac · 2006-02-25 13:58:36

You must have missed Cerebral's post. Initrd doesn't make anything bigger, or more bloated. It just changes how and when things get loaded. The fact that you didn't see them before and now you do with lsmod is inconcequential. You had to "strip down" the kernel to get rid of these before, now you have to strip down initrd to get the same effect.

I really agree with your paragraph there about an additive process. I think that holds very true to what arch is, but you must realize that the kernel will always have to be a subtractive process because otherwise people wouldn't have systems that boot. You can't start with a stripped down kernel, otherwise no one's machine would work because they would be missing every module they need. The only way any precompiled distro can attempt to distribute kernels is by starting them bloated, and informing the user on how to customize from there. Remember how you were saying you would have to edit the .config? That is contradictory to your additive way of arch by being subtractive itself.

And Xorg7 was just released in a poor fashion. I think had their been more warning, everyone would have had a happier experience. Note how Xorg7's modularity fits the additive process

Realize that this is a 'bleeding edge' distro. There is a high chance that things like Xorg7 will come out fast, just like everything else. Don't run pacman -Syu if you don't have time to fix what might break. And I think we should definitely get the ball rolling on initramfs. Of course you could always just use the archck kernel.

beniro · 2006-02-25 14:37:23

Hey, kleptophobiac (me too! ). I agree that Arch is about simplicity, but also it's about staying bleeding edge. I think Arch is still being true to its values, it's just that Linux is a different beast than it was when Arch came to be almost 4 years ago.

If you don't like the idea of staying on the bleeding edge, try hanging back a bit on doing big upgrades until the community gets it straightened out.

In answer to your question, though. I support the switches to udev,initrd and xorg7. udev and xorg7 are leaps forward, IMO (xorg7 hasn't shown fruit yet, but it will). initrd, OTOH, is still unproven. I'm interested in initramfs, though...probably another wierd upgrade.

anykey · 2006-02-25 15:02:42

So, "being bleeding edge" is basically the excuse of all shortcomings and hardships, the answer to all criticism... the standard answer to every rambling... the end-all be-all defense of all design decision...

seems pretty much single-minded to me. Sorry to be this frank.

arooaroo · 2006-02-25 15:22:56

codemac wrote:

And Xorg7 was just released in a poor fashion. I think had their been more warning, everyone would have had a happier experience. Note how Xorg7's modularity fits the additive process

I agree. AL had to keep up to date with this package. However, as xorg doesn't really add anything new, feature-wise, I reckon it could have been kept back in testing for longer. Everything seems so chaotic about this package that I have no intention to upgrade for a good while.

tomk · 2006-02-25 15:27:44

There was an excellent post from Judd at the height of the initrd debate, which I haven't the time to find right now, explaining why he wanted to do it and more broadly, giving us his ideas on Arch and what he wants to be able to do with it. I'm sure it can be found with enough digging, but the larger point that I got from it is that, regardless what any of us think "Arch is all about", it's first and foremost his vision of things that counts, along with the dev team. AFAICS, there are three ways Arch users can handle the decisions they make:

1. Jump right in, test stuff, report bugs, use the tools provided, and support the change process;
2. Resist the change, and attempt to influence it in a way that they believe is better, for them and/or for the larger community;
3. Evaluate, or re-evaluate, alternative distros in the light of Arch changes, to see if they would now better suit their needs.

For those who choose to stay with Arch, I believe the first option is the only way to go.

kleptophobiac · 2006-02-25 19:49:41

The jury is still out on xorg7. It would have to be done sooner or later as software packages came to only support new versions of X. I think, however, more time in testing was warranted since it caused so much heartache for so many people. It didn't really add any new features and so could've been held back.

And yes, it does support the additive process. Good call.

As for kernels being a subtractive process by nature. That's true, but then there's a question of how much do you really need to add? A kernel needs enough bloat to boot, but no more than that. That's usually means of communicating to mass storage, ala SCSI or IDE or SATA. There's nothing wrong with having separate kernels for these fundamentally different ways of getting to mass storage. Nobody was all that confused by the IDE/SCSI kernel split. It also meant a minimum amount of stuff that needed to be subtracted. Perhaps it would be useful to have the -EXOTIC kernel for those that want encrypted root filesystems or other such oddities. Hopefully users pursuing those offbeat goals will be knowledgeable enough to deal with the strangeness associated with their strange setups.

As for arch being bleeding edge.... yes and no. It doesn't make sense to have the newest packages for the sake of having the newest packages. It makes sense to have newer packages to try and maximize features and stability. Just having bleeding edge everything is a recipe for headaches, which is counter to the "simple is better" credo of the arch userbase. Minimalism doesn't beget trendiness.

tpowa · 2006-02-25 20:20:50

the initrd has so many benfits to us, if you don'T like it ok but initramfs is only the successor, more powerfull, phrakture is working on it and i think it will be a good successor.
the kernel this small is a real advantage you can just remove modules you don'T need, there are circumstances when kernels are not booting because of having scsi modules loaded on sata systems etc.
to maintain one kernel package is also a big advantage.
for the rest please dig in the forum / blog etc. there we stated everything why this is great.

Mr.Elendig · 2006-02-26 15:27:44

I belive that many of the troubles with the xorg R7 update woud be solved if people actualy took a look at the news befor they did a pacman -Syu. Most of the questions that have been asked in #archlinux was well described in the wiki. (ofcause the server downtime didn't exacly happen at a "good" moment)

If posible, it woud be nice if there was a warning in packman whenever it's a major upgrade that requiers some action from the user. (prefably before the package is installed, but the way it's done for the kernel package works too)

hotsauce · 2006-02-26 15:36:13

Mr.Elendig wrote:

it woud be nice if there was a warning in packman whenever it's a major upgrade that requiers some action from the user.

I have to second that, very good idea

Arch Linux

#1 2006-02-25 00:25:52

Current path of Arch scaring me

#2 2006-02-25 00:35:21

Re: Current path of Arch scaring me

#3 2006-02-25 00:41:35

Re: Current path of Arch scaring me

#4 2006-02-25 02:58:39

Re: Current path of Arch scaring me

#5 2006-02-25 03:22:52

Re: Current path of Arch scaring me

#6 2006-02-25 03:28:03

Re: Current path of Arch scaring me

#7 2006-02-25 03:36:05

Re: Current path of Arch scaring me

#8 2006-02-25 04:21:42

Re: Current path of Arch scaring me

#9 2006-02-25 04:39:44

Re: Current path of Arch scaring me

#10 2006-02-25 05:51:59

Re: Current path of Arch scaring me

#11 2006-02-25 06:45:53

Re: Current path of Arch scaring me

#12 2006-02-25 09:38:29

Re: Current path of Arch scaring me

#13 2006-02-25 10:27:18

Re: Current path of Arch scaring me

#14 2006-02-25 13:01:32

Re: Current path of Arch scaring me

#15 2006-02-25 13:17:35

Re: Current path of Arch scaring me

#16 2006-02-25 13:19:23

Re: Current path of Arch scaring me

#17 2006-02-25 13:58:36

Re: Current path of Arch scaring me

#18 2006-02-25 14:37:23

Re: Current path of Arch scaring me

#19 2006-02-25 15:02:42

Re: Current path of Arch scaring me

#20 2006-02-25 15:22:56

Re: Current path of Arch scaring me

#21 2006-02-25 15:27:44

Re: Current path of Arch scaring me

#22 2006-02-25 19:49:41

Re: Current path of Arch scaring me

#23 2006-02-25 20:20:50

Re: Current path of Arch scaring me

#24 2006-02-26 15:27:44

Re: Current path of Arch scaring me

#25 2006-02-26 15:36:13

Re: Current path of Arch scaring me

Board footer