A Modern Woody Debian GNU/Linux Installer

An anonymous reader writes "With everyone around talking about how Woody has an outdated installer and lacks some new packages and hardware support, some people feel the urge to get to work. The result? A customized installer. It has a 2.4.26 version kernel, supports XFS, LVM, RAID and various hardware drivers. Comes along with vim, bash, you can even resize partitions using parted and you get postfix as the default MTA. It has two flavours, a business card CD and a miniCD version which will help you install a minimal Debian system or even a X Window desktop."

Comments

Cool

[The Bungi]

Certainly bridges the gap between the Supersize-me 700MB live CDs like Knoppix and the minimalist 10MB vim+kernel+GNU deals.

Should be good even for doing basic partitioning and FS prep before putting in a full distro.

Woody Installer?

[Anonymous Coward]

Who comes up with this stuff! Seriously!

Re:Woody Installer?

[Anonymous Coward]

The same type of people who thought up libpr0n.

Screenshots!

[Profane MuthaFucka]

Reproduced here:

#apt-get update; apt-get upgrade

Now what's so fucking hard about that?

Re:Screenshots!

[Anonymous Coward]

Troll? What the hell are the moderators smoking today? That's funny. Mod it up!

Re:Screenshots!

[Anonymous Coward]

Finding out that the command was "apt-get". Why not "Upgrade System" ??? What is the point of the stupid cryptic command lines? I LIKE command line interfaces! But I want them to make some sense!

Re:Screenshots!

[Anonymous Coward]

Yeah! Like "ls" and "cat". Not that cryptic stuff like "dir".

Re:Screenshots!

[Rysc]

# ln -s /usr/bin/apt-get ~/bin/Upgrade\ System
# UpgradeSystem install mozilla
# Upgrade\ System install mozilla-firefox
Reading Package Lists... Done
Building Dependency Tree... Done
The following extra packages will be installed:
mozilla-firefox-dom-inspector
Suggested packages:
latex-xft-fonts
2 packages upgraded, 0 newly installed, 0 to remove and 2846 not upgraded.
Need to get 10.5MB of archives.
After unpacking 397kB disk space will be freed.
Do you want to continue? [Y/n]

Of course all of the on-screen text refers to apt, but it can be done.

In serious reply, the reason it's not called Upgrade System is that that's not exactly what it does, and more importantly it's not the only way to do it. APT just happens to have been adopted by Debian. Why say Mozilla and not Web Browser? Because it's not the only web browser. Why say Mozilla and not Internet? Because the web is not the internet. Why say apt-get and not Upgrade System? Because apt is the method being used, and that's not precisely what it does.

Why be accurate instead of "friendly"? Because we tried friendly, and it's better to be accurate if that is the choice.

Writing an installer? Make it portable. Please.

[dotz]

What I'd really like to see would be something universal and portable:

• divide the installer code wisely, you will have UI part and the installer part (that does actual work, system-dependent) separated.
• GUI? It shouldn't be that hard. XFree86 for graphics installer are the same everywhere;
• Want to instal via serial port? No problem, just add another user interface module
• High-level language, not C. Sorry, C programs just need too much time to debug, and I don't see where would you have any benefits of using C in case of installer (installation process always takes time, and it mostly depends on HDD or network throughput)
• There are some OS, that lack an easy graphics mode installer. They could benefit

That would of course need a few megabytes of RAM and an isntallation CD. Is anyone booting off from floppy disks anyway? (what's a floppy disk, BTW?)

Re:Writing an installer? Make it portable. Please.

[Trepalium]

Is anyone booting off from floppy disks anyway? (what's a floppy disk, BTW?)

It's the thing your BIOS emulates when booting from most el torito bootable CD-ROMs.

Re:Writing an installer? Make it portable. Please.

[Anonymous Coward]

So what's stopping you? You don't have to write any code at all if you can get involved with one of the installation projects in a testing or advisory role.

Re:Writing an installer? Make it portable. Please.

[dotz]

Terrible lack of time. Sorry! I thought I share such idea instead.

Re:Writing an installer? Make it portable. Please.

[Elivs]

I think what you want is the new sarge installer that is at RC1 at the momemnt.

Its modular to support all the things you want, and supports 10 archetures at this stage. Being modular should allow people to: script it, put a GUI on it, hardware autodection modules (already done), multiple boot methods (PXE,USB mass-storage,CD-rom...)

Elivs

Re:Writing an installer? Make it portable. Please.

[shaitand]

Just one anal point here. C IS a high level language. A high level language is any language which uses commands and functions rather than instructions. You must be confusing VERY high level languages with high level languages. Most (but not all) very high level languages are interpreted.

High and very high level languages consist of most languages. Low level languages are short in supply, mostly being limited to assembler and machine code.

Re:Writing an installer? Make it portable. Please.

[BerntB]

C IS a high level language. A high level language is any language which uses commands and functions rather than instructions.

"commands and functions" -- a Macro assembler fulfills that.

For fun, Google for PDP-10 assembly language and compare it with that era's C (pre-'void'!).

I'm not at all certain which was the easiest and neatest of those two.

Re:Writing an installer? Make it portable. Please.

[shaitand]

Well it does if you take the quote out of context instead of reading the whole sentence and realizing that "rather than instructions" was a requirement as well.

Really though, a high level language which isn't even easier to use is no less a high level language. Just a shitty one ;)

Re:Writing an installer? Make it portable. Please.

[BerntB]

"rather than instructions" was a requirement as well.

A macro assembler can do quite complex things.

Re:Writing an installer? Make it portable. Please.

[shaitand]

What does that have to do with high level languages not using cpu instructions directly?

And no, embedded asm or hacks to make it happen doesn't count.

Re:Writing an installer? Make it portable. Please.

[BerntB]

You wrote:

"rather than instructions" was a requirement as well.

And I explained (again) that the connection between assemblers macros and their instructions aren't necessarily direct; some truly weird && / || wonderful things can be done. Which are far removed from simple "one line -- one instruction".

Now you wrote:

What does that have to do with high level languages not using cpu instructions directly?

what is the point?

Do you only know about the disgusting x86 architecture and can't understand how neat it can be done if you have lots of registers?

Re:Writing an installer? Make it portable. Please.

[shaitand]

Somehow I think we are debating but neither of us are talking about the same thing??

I'm discusing whether C is a high level language or not. Not whether high or low level languages are good and which is superior to the other. If you want flexibility, control, speed, or efficientcy then a low level language is the way to go. If you want portability than a high level language is the way to go.

All of the ease of use and abrastraction in a high level language can be built with a low level language. Going lower level INCREASES capabilities, it certainly doesn't reduce them.

As for the instructions, somehow I think your reading my statements backwards? I'm not claiming that a high level language is high because you can use it without using CPU instructions directly, or that a low level language is low because you cannot code in it WITHOUT using cpu instructions.

I'm saying a high level language is high because you CANNOT code cpu instructions directly. Again the reverse does not make a language high level, being able to code without using cpu instructions.

In a high level language everything MUST be done through a layer of abstraction or embedding of a low level language. Although in C's case you can use very dirty hacks to coerce cpu instructions into ram, proper C usage doesn't support this.

Re:Writing an installer? Make it portable. Please.

[CentrX]

C is often called a mid-level language because it's lower than pretty much every other language except assembly. You know there is something between the words "high" and "low".

Re:Writing an installer? Make it portable. Please.

[dotz]

Elivs, by different architectures I meant also non-Debian _and_ non-Linux operating systems. Cries for graphics installer appear on freebsd-advocacy@ mailing list from time to time (too often, if you ask me or other advocacy@ regulars). Of course - it is not that needed, but if someone did it, the project for sure wouldn't suffer from this. This also counts for OpenBSD and NetBSD... anyway, I don't think that there are as big chances for autodetecting stuff on BSDs, like they are on Linux. Perhaps I'm wrong, perhaps not.

Re:Writing an installer? Make it portable. Please.

[dotz]

Yes, indeed. Except, that the most often used format is 2.88 MB floppy - I haven't seen such floppy drive anywhere around. Are they rare/unused, is that the same in other countries - or were they popular in some area, some time ago, at least?

Re:Writing an installer? Make it portable. Please.

[BerntB]

I'm discusing whether C is a high level language or not.

Oh, that answer is obvious:
It depends. :-)

You can do a very simple translation to any assembly language (with a subroutine library) for most of early C. Some stuff like expressions (a+b/c) aren't usually supported in assemblers and register allocation needs to be worked at, too. (To make it effective and fast is another thing, of course.)

C can be seen as a good macro assembler that is machine independent -- it has no concepts that aren't easily translated into assembler. Which makes it quite a neat creature.

A high level language is something that implements abstract models, usually from mathematics, but I don't really see any reason to exclude low level stuff because of that (car and cdr in lisp, for instance). It's usually not a good idea.

I think you overrate access to machine instructions as a definition of high/low level. I've seen parsing and graphics algorithms that generate code for a simple processor that is then emulated. All that could easily be implemented in a high level language.

Re:Writing an installer? Make it portable. Please.

[Elivs]

The debian installer is aimed at not just different hardware ports, but also different kernels including netBSD, FreeBSD, and Hurd. (http://www.debian.org/ports/#nonlinux) While these are not yet "stable" ports the installer certainly tragets them in its design. Acknowledgeing that the installer only target debian versions of these kernels.

If you view an installer in the most basic sense, all it does is partition the disk, unarchives a few files, and install a boot loader. The boot media (floppy/net), kernel (bsd/linux), and GUI (gtk/ncurses) used to do this can all vary and be unrelated to the actual system being installed. I think this is the sort of thing you where suggesting and its the aim of the sarge installer. The difficulty comes in making the installer flexable and extensible enough that you don't need to rewrite the whole thing for every different situation.

With the new sarge installer the flexability come from modularity. Each part of the installer is a ".udebs" package (micro-debian packages) that resides in the debian archive/bug-tracking system like other packages, complete with dependences etc. Thus someone building an image for an installation disk can pick and choose which modules are include depending on their needs. They can even write there own module if they want. For a fictious example: they might want to use a QT GUI rather than GTK, use hardware autodectection for sparc rather than i386, install Hurd rather than linux, and choose GRUB as the boot loader.

It will be interesting to see how well the new design actually achives this and its other goals. But, like most things, in debain flexability is considered a very design goal.

Elivs
(disclaimer: I'm not a debian developer, but I've been using debian for 5yrs and try to follow many of these sorts of dicussions)

A Modern Woody?

[warpSpeed]

Bevis and Butthead
ahuhhuhhhuhh... he said woody.
/Bevis and Butthead

Mod parent up

[Anonymous Coward]

Ninnle is the one distro that I feel most comfortable with. It's easy to install, configure, and maintain. People don't know what they're missing.

Re:Mod parent up

[Anonymous Coward]

How can this be insightful while the parent is offtopic?

Anaconda

[Guspaz]

I never understood why Debian doesn't simply use Anaconda. It's opensource, portable to Debian, and everybody who uses it seems generally pleased with it.

Re:Anaconda

[Anonymous Coward]

There are lots of installers that debian could have had which are much nicer than the current one in terms of user experience. The Progeny installer, which was developed for debian, is open-source, and has been around for a long time, is the most obvious example. The Debian project has rejected all of them in the name of portability, but they won't expend the effort to port them mainly for religious reasons. They have completely redesigned the installer for the latest release; it has a wacky, modular structure designed (I guess) to make it more portable and easier to build, but it still has the same lame text-based interface as before (in some ways it is even less intuitive actually, because it tries to do many more things automatically and then gives less-than-comprehensible feedback when it inevitably fails with your hardware). They claim that this will make it "easy" to write an easier-to-use front end eventually, but given the history of these things I don't see that happening.

Anaconda: Not Interesting (!!!)

[hummassa]

Anaconda only works in x86 (and sparc?), and debian has to install in 11 archs. Understand now?

Re:Anaconda: Not Interesting (!!!)

[Anonymous Coward]

Nope. Still don't understand. What's easier - porting an installer that is already cross-platform to more architectures, or writing a new one from scratch?

Re:Anaconda: Not Interesting (!!!)

[Rysc]

Will it work if my monitor is a printer or a real teletype? No? Then it's not as good as the Debian installer. You know, the once that doens't assume you'll be able to fire X right up? That one.

Re:Anaconda: Not Interesting (!!!)

[Anonymous Coward]

doens't assume you'll be able to fire X right up?

And more importantly, doesn't assume you'll want to fire X right up.

X has no place on a firewall, any server (except an X server, of course), or any displayless systems.

Re:Anaconda: Not Interesting (!!!)

[Guspaz]

Anaconda is written mostly in python. It should be easily portable to any arch. Somebody just has to take the effort to do it. And a major distribution (should) have the resources to pull that off.

Turns out there's already a port of Anaconda for Debian, though it's i386 only:

http://platform.progeny.com/anaconda/index.html

Re:Anaconda: Not Interesting (!!!)

[Guspaz]

How many people run Linux with a printer/teletype as their primary output device? Not enough to care about. A few odd people shouldn't hold back a distribution from getting a usable installer.

Besides, Anaconda has a text installer... I'm sure if one of the three people who need teletype support was desperate, they could create some sort of an output filter to make the text installer support teletype compatible output.

Re:Anaconda: Not Interesting (!!!)

[smurf1974]

More people than you think is installing debian in strange ways.

I recently installed debian on a MIPS based Set Top Box, via a nfs mounted home directory and a serial console. I mainly did this because the linux system that came with it from the chip vendor was a screwed up redhat port and I liked Debian better when developing for the platform.

Re:Anaconda: Not Interesting (!!!)

[Guspaz]

Ah, but you said the magic word, "Serial Console". Anaconda's text-mode installer wouldn't work well with a teletype because a teletype can't go back and modify what's already been printed, making it impossible to display a text-based GUI.

However, a serial console has no such limitations. You could have used Anaconda's text-based mode for your installation without trouble, assuming it had been ported to your platform (MIPS).

I was semi-serious when I suggested an output filter in a previous comment. If a filter was written that took as input the Anaconda text-based output, and turned it into linear questions-and-answers, then Anaconda would work on a teletype just fine.

Modern Woody?

[GregChant]

Isn't this a contradiction? The whole point of a frozen release is that nothing changes. It's what allows Debian to say its system is remarkably stable.

If you change packages or change the kernel, it isn't Woody (the reknowned stable version) anymore, and instead becomes Sid (the more up-to-date, but labeled unstable version).

I wonder what the Debian overlords are going to do with this...

Yeah but...

[Oriumpor]

Why woody ticked me off about a week ago:

1. 2.2 kernel
2. Hardware Support
3. 2.2 kernel
4. See #1, #2

Solution:
1. Install Fedora Core
2. install apt
3. be happy.

Re:Yeah but...

[martinde]

> 1. 2.2 kernel

2. Slap forehead, keep woody install CD in CD-Rom drive, reboot, read help by pressing F-whatever it says.
3. Instead of hitting return to boot, follow the directions you found in the help and do "bf24" at the prompt to boot into 2.4
4. Hopefully that gets you going...

I'm not saying woody is perfect - i've had to install PCI ethernet cards too many times because the default kernel won't do modern integrated ones... But it does support 2.4.18 which is much better than 2.2.

twin

[samjam]

Use twin - its like screen but can back-connect to an X-server!

I modified the new sarge root disks so I could do remote installations without being at the console.

Sam

Woody's "up to date"

[Ianoo]

How long has kernel 2.6 been out now? Hence, why is this "up-to-date" installer stuck at 2.4?

Re:Woody's "up to date"

[Anonymous Coward]

How long has kernel 2.6 been out now? Hence, why is this "up-to-date" installer stuck at 2.4?

Because it's installing Woody. And the most recent kernel in Woody is 2.4.

Please read up on the Debian release system before making clueless comments.

My Experience

[magnum3065]

1. Install Fedora Core 2
2. Install apt
3. Realize I have to add a bunch of extra repositories to get a half-way decent selection of packages
4. Still can't find packages for lots of stuff I had in Debian
5. Overlaps and conflicts between packages in different repositories causes havoc everytime I try to upgrade
6. Not so happy

Re:My Experience

[metamatic]

Same here, and then Fedora crapped all over itself and I had to reinstall.

multiple architectures and reduced bandwidth

[jdowland]

The authors should provide a jigdo file. Then, the installer image could be built by hammering the FTP mirrors rather than their own space.

In addition, it would be possible to build such a boot image for non-x86 architectures; the reason of course, why none of these `better' approaches have replaced the current debian installer.

Question about using non-standard installers

[astrashe]

I hope this isn't a dumb question, but...

Do non-standard installers have an effect on security updates?

I've wondered about that with livecd distros that can set up debian systems on a hard disk. If they draw their packages from standard sources, you'd have to figure that the updates would come through ok.

But what about the things the installer itself sets up? Does it all come from packages that will be updated, or does some of the system come from files on the install media that aren't covered by package update?

Absolutely hopeless.

[multriha]

A quick check of the page shows no screenshots. OSS without screenshots is just not competitive.

The Debian team has already done this

[peripatetic_bum]

Not sure why they started this since Debian has a project called Debian Installer. I have used on some very modern boards and it has really done an amazing job in detecting all my software and running a 2.4 kernel.

Please check out Debian Installer. I think you will be plesantly pleased

Not quite the same

[Poor College Student]

The new Debian installer does Debian testing. The article mentions one for woody.

That said, I do agree that the new installer is quite straightforward and did detect all my hardware, and I agree that this new installer isn't going to be a longterm solution as the new Debian installer will be the default when sarge becomes the next stable.

Re:Not quite the same

[cortana]

Before installing the base system you are prompted which release you want to install: stable, testing, unstable. So AFAIK you can use debian-installer tc1 to install Woody.

Hahah

[lorcha]

The authors should provide a jigdo file.

I wish I had mod points right now so I could mod you as funny. Jigdo. Hahah. Funny.

Debian is dying

[Anonymous Coward]

fact> debian is dying
no new releases...
they still have the linux kernels 2.2.x and 2.4.x
archaic installer
it's full of bugs / the most buggy and unsecure open source os. linuxsecurity.com/advisories

Debian is dying

[Anonymous Coward]

Debian doesn't use anaconda because...
Debian is dying.
Next debian release: never
(Score:5, Interesting)