wish list, 2.6


Forum: DSL Ideas and Suggestions
Topic: wish list, 2.6
started by: jpeters

Posted by jpeters on Nov. 22 2007,10:38
It might be fun to have a stripped down dsl with 2.6 kernel just for experimenting; say no apps other then a terminal, very minimal desktop, etc.  (i.e., device centric).
Posted by john.martzouco on Nov. 22 2007,12:37
I think this is a great idea!
Posted by roberts on Nov. 22 2007,16:15
It is more than just the kernel that makes for a distribution. You would want updated or matching libraries and other support files.

When John started the DSL project, he used a tiny 22MB knx business card Linux distro. He built tinyX the k-drive server and selected and added applications. This was the base for DSL up to 0.5.3.1.

Starting with v0.6, I actually stripped the full Knoppix v3.3 then added the DSL layer of kdrive X server, the apps, and my custom code.

I again stripped Knoppix 3.4 for DSL v0.8.0, and once again went throught the process of moving kdrive, apps, and all my custom DSL code.

The point being made is that the Linux foundation should be a matched set of kernel/modules, libraries and support utilites.

It is too much work to strip Knoppix these days. There is the T2 build system, there is LFS to also build a base, and of course "standing on the shoulders of others". Using an existing base, e.g., Austrumi & Puppy are slackware based, Pudge and 40+ other are now Ubuntu based.

Building from scratch wether T2 or LFS is fine when building your own system. The real challenge comes when making the distro public. Then suddenly you are faced with all kinds of hardware that you personally don't own. Building modules and often support files for such is most difficult. No access. No debug. That is the reason so many distro remaster. By using a proven known base, the distro maintainer can concentrate on what they like to do, or what they do best. Adding custom code and making the distro unique is what made DSL.

"Of course some distros are really nothing more than swapping a window manager, or just a slightly different collections of base applications."

The above statment is not mine, I hear so often when I walk the floors of any Linux World or Linux related conference.

To me a minimum useful base for a 2.6 foundation would be,  full debian capable, unionfs/aufs, gtk2, and firefox web2.0/flash9 capable. This is my goal and what I have been working on, on the back burner. DSL is down to one developer and I have had my hands full with 4.x. But 4.x is becomming quite stable and once 4.1 is released I will likely pick up on my 2.6 project and see where is goes.



Posted by Juanito on Nov. 22 2007,17:37
Quote
But 4.x is becomming quite stable and once 4.1 is released I will likely pick up on my 2.6 project and see where is goes.

- Great, I'm looking forward to it  :)

Posted by jpeters on Nov. 22 2007,19:45
Quote (Juanito @ Nov. 22 2007,12:37)
Quote
But 4.x is becomming quite stable and once 4.1 is released I will likely pick up on my 2.6 project and see where is goes.

- Great, I'm looking forward to it  :)

Yes, and once again thanks for all the outstanding developments in the present version of DSL.

Posted by dougvega on Nov. 28 2007,23:18
this is great news I allways wanted to install the Kdrive in a true debian distro
but lacked the know how .
I hope your new development keeps the tiny X that makes DSL so fast
best wishes in this new undertaking

Posted by roberts on Dec. 02 2007,23:16
Eariler in this thread I wrote
Quote
...once 4.1 is released I will likely pick up on my 2.6 project and see where is goes.


I released DSL v4.1 today. So I am now going to turn my attention to this new project.

I do not want to create "a me too" distro. I would let the project die before I let this happen.  I do not want to be included with < The Convenient Fiction About Distributions >

I do not have the time, inclination, or desire to begin from scratch. As I ranted about in other posts, I get no joy compiling the creative work of others. On the other hand, I have no problem giving credit where credit is due. Doing so (standing on the shoulders of others) is what freedom in software is all about.

I plan to start by putting together a very minimal base. I will be using much of my lua, murgaLua and Fltk as possible. This may look like a minimal version (tiny core) of DSL, but it will not be. It will not even support the same structure. No gtk1, no cloop, no MyDSL extensions (at least as we know them today). Newer more robust and capable support systems makes this possible.

This will likely be in 'development' longer than any other version of DSL. John did the first one, a 2.4.22 kernel, based on the business card knx distro. I created the 2.4.24, 2.4.26 (based on stripping  corresponding Knoppix versions) and finally 2.4.31 (by compiling and updating)..

I will likely call upon the community to become more involved when it comes to support of devices and hardware that I cannot afford or have access.

I wish to continue the DSL tradition (actually a real Unix mantra) that small is beautiful.Many in the past have asked for a tiny core. Well this will be such.

I would not suggest that anyone wait till "its ready". That date is undetermined.

DSL currently offers 10 Editions, five in the 3.x series, a stable, familiar, robust, and well documented system.
If you prefer single click application launchers and/or application menu driven, then 3.x should be your choice.

DSL offer the same 5 editions in the 4.x line. If you prefer folder, documents drag-n-drop, data aware, data centric system then 4.x is for you.

I will continue to support the latest releases of both series of DSL.

I just feel that I need to begin the process to offer something 2.6 based. Something to build upon and not something to strip down. DSL-N was a project that began to try to strip Knoppix 4.02 but Knoppix was too much of a moving target and there was still too much that I wanted to accomplish with the 2.4 kernel based systems.  And of course I was very busy writing a book. DSL-N suffered from both lack of community interest and lack of my available time and attention. Support systems technology has also moved on and imporved so that it is not a viable option to continue DSL-N.

It may likely be a long road. But it is time.

Posted by Juanito on Dec. 03 2007,12:49
Great - I'll buy a new usb stick in anticipation of the testing :)
Posted by john.martzouco on Dec. 03 2007,15:10
Robert,

I tip my hat off to you for taking on the new challenge.  I'm very much looking forward to the 2.6 kernel based tiny OS and know that it will be of the highest quality because of the man who's building it.

With my utmost respect,
John

Posted by roberts on Dec. 14 2007,00:58
Quote (jpeters @ Nov. 22 2007,02:38)
It might be fun to have a stripped down dsl with 2.6 kernel just for experimenting; say no apps other then a terminal, very minimal desktop, etc.  (i.e., device centric).

I have such with a couple of images. But it seems like trying to be Debian is a catch-22. If I am true Debian, adding X, will bloat us immensely. If I sneak in a tinyX, which I have, then I can, of course, go non-Debian to add other X apps. But if you try to add an X app via Debian, here comes the bloat. Perhaps that is why, or so it seems, that every small 2.6 distro have choosen slackware as it base. Debian dependencies are great on a real hard drive, but when building something very small, I am not so sure.

Perhaps the approach to take is to "sneak in", i.e., non-Debian, a minimum of tinyX, a tiny WM, an xterm. Lets this tiny core be under the Debian radar. Then if a user wants to install true Debian apps and does not mind the size, they can. The downsize is that it is not very useful. Oh well, just thinking out loud. Care to join in?

Posted by dougvega on Dec. 14 2007,01:19
There was a post on the smallbuntu forum on how to ad tiny X to ubuntu , I could not get it to work ; to technical for me
but I read some users made work
perhaps building a small Distro truly Debian Base and then let the user decide whether to install the Kdrive is the way to go an please everybody
I'll se if I can find the link to the Smallbuntu topic and post it later.

Posted by Juanito on Dec. 14 2007,08:37
For me it doesn't neccessarily need to be debian based, but if it is, then I think it would be good that all the libs/apps come from the same version of debian to avoid the whole broken depedency story.

Maybe it would be better non-debian to have the freedom to chose the best/fastest/smallest core. We would also have exactly the right headers to build on.

Posted by curaga on Dec. 14 2007,20:43
With a coherent core, we could archieve extensions working better together, right versions of headers in compiling extensions, and right version man pages, for example..
Posted by WDef on Dec. 15 2007,11:51
Quote
continue the DSL tradition (actually a real Unix mantra)


Amen to that Robert.  We can get our dose of bloat anywhere else.

Re: that article about the 'convenient myth' - I think it's a good thing that there is convergence between the main distros and the LSB.  Much of the divergence for desktop generic users only fragments the linux world.

There will always be specialized distros for subsets of users and specific applications, though.  DSL is one of those.

One thing the article fails to address is the very different approaches of the 'stability is paramount' camp (Debian and Slackware) versus the 'have the very latest' camps (eg Fedora and Gentoo).  These aren't converging as much and they both have their uses - the latter are good for trying out the newest apps etc, the former are great for a low-threat, non-hair-pulling experience.

Posted by Juanito on Jan. 06 2008,12:31
Quote
Great - I'll buy a new usb stick in anticipation of the testing

- OK, so I bought myself a new usb stick - any update on possible alpha release dates  :)

Posted by roberts on Jan. 06 2008,16:00
As you can imagine, I have been very busy with the launch of my retooled v4.x series. The last several releases have been to stablize the series based on community feedback. With over 750 extensions and all the incarnations that exist within I needed feedback to better support them.

I have been reviewing several frameworks and have protoyped a couple. Currently I am leaning away from Debian and more towards Slackware.

On a personal note, I am struggling with my Muscular Dystrophy and its effects on my vision due to extreme ptosis. I will be scheduling surgery to try to correct. Don't know how long that will put me down.

Posted by muddywaters on Jan. 06 2008,16:46
Quote (roberts @ Dec. 02 2007,18:16)
I will likely call upon the community to become more involved when it comes to support of devices and hardware that I cannot afford or have access.

Quote
I will likely call upon the community to become more involved when it comes to support of devices and hardware that I cannot afford or have access.


Any devices and hardware in particular?  Hopefully not too pricey?

The thread is called wishlist afterall.  :)

Posted by Juanito on Jan. 06 2008,17:15
Quote
On a personal note, I am struggling with my Muscular Dystrophy and its effects on my vision due to extreme ptosis. I will be scheduling surgery to try to correct. Don't know how long that will put me down.

- best of luck with that.

Posted by roberts on Jan. 23 2008,22:33
Update:

I have been prototying various builds and want to share some of my thoughts with the community.
I am considering a non unionfs kernel 2.6 system for the next major DSL version..
Currently I am booting a prototype 36MB Knoppix 5.1 kernel and modules without unionfs.
It boots to JWM/DFM with all my Lua and custom Bash scripts. Other than that it is currently app-less.

DSL up to and including v2.4 did not have unionfs.

Community member clacker first created a unionfs.dsl extension to add unionfs to DSL.
So that could still be an option for the new version. There is also a fuse based overlay available.

Many of the DSL community seem to  shun unionfs. It is too buggy even in newer releases.

To quote from the:
< February 2007 Linux Storage and Filesystem Workshop >:
"Some of these issues stem from the fact that the Linux Kernel has not been designed with stackable file systems in mind.

In hindsight, I sometimes wish I had not included unionfs into DSL. I am so glad that I have maintained a "legacy" style boot. As it is far more reliable.

Current count in the repostory is
53 unc
182 uci
463 dsl
237 tar.gz

There has been alot of recent builds of uci. Very little new unc.

My original concept of self-contained applications via iso/cloop was spot on.
Later, I see Klik using the same concept except they use cramfs.
Then I see zeroinstall being developed by rox group.
I see Gobol also trying to promote "chrooted build" self-contained applications.
PCBSD has PBI, similiar concept. And Macintosh uses self contained application folders.

All of these validate my early idea. My first DSL extension was called a .ci (compressed image)
I released .ci and .tar.gz as the first example myDSL extensions to the DSL community.

I actually was against the mkwriteable script used in DSL. It was only because of the challenges of building self contained applications that I added mkwriteable into DSL. Doing so caused a flood of new .dsl extensions.

I am thinking to wait out the unionfs/aufs/posixovl for better support. But I am also still wanting to promote my original concept of uci and see how it fares with the new knowledge that our community has built upon.

Have you see the gymnastics in the new knoppix-halt, knoppix-reboot scripts to try to handle busy inodes caused by stackable filesystems? It was the reason that I disabled tyring to un-mount a unc in DSL.

I guess I could delay a new build of DSL until stackable filesystems are better supported.  But I am thinking a non-unionfs DSL might be a success. It would boot with mkwritable defaulted so we would still have all the current capabilities of a "legacy" booted DSL.

Comments

-- Robert

Posted by lucky13 on Jan. 23 2008,22:55
I just sent you a note updating my progress with Knoppix 5.1. Or lack of it. I did update the toolchain, though, and the Debian kernel (2.6.23.1) is a lot faster to download than to compile on this old wreck.

I wouldn't object to moving apps from the base to MyDSL. That was one of the things I lobbied for in DSL 4 -- I'm openly very biased in that direction. Doing so now would allow you room for more modules so both legacy and new hardware can be as evenly supported as possible. Your kernel is already way smaller than the Knoppix 5.1 one:
Code Sample
$ du --si --max-depth=0 /lib/modules/2.6.19
69M     /lib/modules/2.6.19


And this is the update I did at Debian earlier today:
Code Sample
$ du --si --max-depth=0 /lib/modules/2.6.23-1-686/
55M     /lib/modules/2.6.23-1-686/


My first 2.6.23.14 (gcc 3.3.4 against the base c lib in DSL) with *every* possible module was just over 80 MB. But I know you won't include bluetooth and irda in the base.

Moving more apps to MyDSL would also afford users more choice -- GTK1 versus GTK2 Firefox, Opera, etc. -- without duplicating what's already in the base.

Posted by Juanito on Jan. 24 2008,04:10
Quote
In hindsight, I sometimes wish I had not included unionfs into DSL. I am so glad that I have maintained a "legacy" style boot. As it is far more reliable.

- I always use a legacy boot to compile. Many times when a compile fails on a standard boot "weird things" happen and I cannot delete files/directories. For sure this could be due to many things (including my mistakes) but it doesn't happen with a legacy boot.

Quote
There has been alot of recent builds of uci. Very little new unc.

- I think this is because the "community" has learnt how to make uci extensions, perhaps due to the excellent dsl book. On the other hand, I suspect a lot of users were making dsl/unc extensions from deb woody packages which are are no longer easily available.

I believe the new 2.6.x dsl should go with a "legacy" core with as many modules available as can fit/makes sense to ensure maximum compatibility and almost no apps (perhaps this is selfish, but I rarely use many of the base apps) in the base system. Users could then use uci extensions to suit their needs.

Posted by roberts on Jan. 24 2008,06:36
Juanito wrote:
Quote
believe the new 2.6.x dsl should go with a "legacy" core with as many modules available as can fit/makes sense to ensure maximum compatibility and almost no apps (perhaps this is selfish, but I rarely use many of the base apps) in the base system. Users could then use uci extensions to suit their needs.

This is exactly what I would like to promote!

mikshaw, once made a graphic for DSL that said "I love Uci" a spin on "I Love Lucy". That does provide the most stable and easily used  (mounted/umounted at will) extensions.

BTW: I booted my 36MB iso then loaded gtk2 extension, downloaded the lastest Firefox then browsed to youtube. Easily downloaded and installed flash9. Ran several youtube videos. Nothing crashed. Worked flawlessly.

Posted by Juanito on Jan. 24 2008,07:09
Quote
BTW: I booted my 36MB iso then loaded gtk2 extension, downloaded the lastest Firefox then browsed to youtube. Easily downloaded and installed flash9. Ran several youtube videos. Nothing crashed. Worked flawlessly.

- excellent (not that I ever used YouTube), when do we get to play with it  :)

Posted by curaga on Jan. 24 2008,08:04
@Juanito: when I compile on DSL, I do a standard boot, but keep the files either on HD or in ram (/ramdisk, so unionfs doesn't get to mess with it)
Haven't had problems.

Posted by jpeters on Jan. 24 2008,08:27
Quote (roberts @ Jan. 24 2008,01:36)
Easily downloaded and installed flash9. Ran several youtube videos. Nothing crashed. Worked flawlessly.

I'm not concerned about the youtube videos (which work with flash 7 anyway), but AM interested in being able to access my brokerage accounts, which are switching to flash 9 for loggins.....(e.g, TradeKing, as of Feb 9th)

I  thought unionfs were more efficient in terms  of RAM; also easier to build. (I gave up trying to figure out why my initial attempts creating UCI's loaded from some directories but not others....nobody else was able to solve it either, as I recall).

Posted by roberts on Jan. 24 2008,16:26
Quote (jpeters @ Jan. 24 2008,00:27)
Quote (roberts @ Jan. 24 2008,01:36)
Easily downloaded and installed flash9. Ran several youtube videos. Nothing crashed. Worked flawlessly.

I'm not concerned about the youtube videos (which work with flash 7 anyway), but AM interested in being able to access my brokerage accounts, which are switching to flash 9 for loggins.....(e.g, TradeKing, as of Feb 9th)

I  thought unionfs were more efficient in terms  of RAM; also easier to build. (I gave up trying to figure out why my initial attempts creating UCI's loaded from some directories but not others....nobody else was able to solve it either, as I recall).

Anybody intending to run a 2.6 kernel system will have to have capable hardware, i.e., much more ram. The kernel alone in 2.6 is twice the size of 2.4.

As for efficiency, legacy boots faster than unionfs in my 2.6, both use ramdisk as storage, unionfs has much "housekeeping to do" (an additional process). Unionfs if desired can still be used as an extension, see unionfs.dsl in the repository.

The youtube flash9 is a simple "acid test", as flash9 in current DSL crashes. I am not an avdanced web user so if you know any other flash9 tests I can try please let me know.

Posted by ^thehatsrule^ on Jan. 24 2008,17:13
Quote (roberts @ Jan. 24 2008,01:36)
BTW: I booted my 36MB iso then loaded gtk2 extension, downloaded the lastest Firefox then browsed to youtube. Easily downloaded and installed flash9. Ran several youtube videos. Nothing crashed. Worked flawlessly.

Was this with sound as well? is ALSA in the core? (or if not, using a wrapper?)

I'm looking forward to this release as well - mainly to see how DSL integrates with the newer kernel... and how it will fit with my computing future.  Is this going to be declared as version 5.x?

Although I still use .dsl's for certain things, most of my applications are .uci's - loaded when I need them (I don't use unionfs).

Also, I think there could be some library conflict that had flash 9 crash for many users... on their system requirements, they don't list a 2.6 kernel (that was posted around here) and they even have an officially supported 2.4.x distro.

Posted by jpeters on Jan. 24 2008,17:45
Quote (roberts @ Jan. 24 2008,11:26)
Unionfs if desired can still be used as an extension, see unionfs.dsl in the repository.

Will they still run as unc's (i.e., cling to their hda locations) or as dsl's? If the latter, we're probably looking at 512M RAM minimums (to be safe).

Posted by roberts on Jan. 24 2008,18:27
It also possible that other core libraries will be included in the iso.
What I am currently running is only a prototype. A proof of concept.

However 512MB is not a bad target given today's standard.

I will not be dropping DSL 4.x. I still use it and will remain using it on  my smaller less capable machines.

Posted by chaostic on Jan. 25 2008,07:08
Quote (roberts @ Jan. 24 2008,13:27)
It also possible that other core libraries will be included in the iso.
What I am currently running is only a prototype. A proof of concept.

However 512MB is not a bad target given today's standard.

I will not be dropping DSL 4.x. I still use it and will remain using it on  my smaller less capable machines.

Ha.
3.x = DSL Classic/Legacy (""Obsolete""/Past "useful" life cycle comps)
4.x = DSL Proper (Near end of Life Cycle ~ Current Life Cycle comps)
5.x = DSL FutureSight (Begining of Life Cycle/New comps)

>_>

From my use of DSL (I stick to 3.x), it is perfect and like 0 bugs for me. So I wouldn't care/need constant updates, as I have no problem changing the systems guts if needed. Maybe just kernel security patches every other month?

And as for DSL Future, how about 200~250 mb? That way, it will take up all of a mini-cd, still keeping it small.

Edit:
Oh wait, 512 = ram. I thought size. :/

Posted by lucky13 on Jan. 25 2008,15:25
Quote
Maybe just kernel security patches every other month?

Aside from hardware support, there's not a reason to upgrade kernels periodically. That's too intensive a process for something that's intended to be used as a live CD or installed in a manner so that the kernel is in a read-only environment.

Posted by chaostic on Jan. 26 2008,13:05
Quote (lucky13 @ Jan. 25 2008,10:25)
Quote
Maybe just kernel security patches every other month?

Aside from hardware support, there's not a reason to upgrade kernels periodically. That's too intensive a process for something that's intended to be used as a live CD or installed in a manner so that the kernel is in a read-only environment.

Hence "Maybe" :p

But maybe more in the line of bug-fix updates to current programs in dsl 3.x, like the supposed fluxbox/x-11 memory leak (fat chance, but I have 10 days uptime and no leak, hence supposed) or the white-page bug in monkey, like will be updated soon.

Posted by jpeters on Jan. 27 2008,05:32
Quote (roberts @ Jan. 24 2008,11:26)
Anybody intending to run a 2.6 kernel system will have to have capable hardware, i.e., much more ram. The kernel alone in 2.6 is twice the size of 2.4.

Okay; just found a rebuilt Dell D600 with a gig of ram on Craig's list (for all of $300....most expensive laptop I've purchased in years! ) for the experiment.  Hey....I've got great expectations....this should be fun, anyway....(I realize there's stiff competition out there in the "damn small" clones ....and am hoping you'll be including X  in the release  :)  )

Posted by Juanito on Jan. 27 2008,05:37
Quote
just found a rebuilt Dell D600 with a gig of ram

- probably the first thing you'll find is that you'll need XFree86 and 855patch/wrap or Xorg72 to get the display working properly...

Posted by jpeters on Jan. 27 2008,05:42
Quote (Juanito @ Jan. 27 2008,00:37)
Quote
just found a rebuilt Dell D600 with a gig of ram

- probably the first thing you'll find is that you'll need XFree86 and 855patch/wrap or Xorg72 to get the display working properly...

It booted up 4.2 no problem... (think I was able to set vga with just xsetup)

EDIT: I checked it out again; boots up at 1024X768; has a Radeon Mobility 9000 M9 video card, and can't get higher res. I have a C600 that only gets 1280X1024, and has a Rage Mobility M3 AGP 2X card (can't get a lower res).  Another C600 has a Radeon Mobility M6  card and gets 1024X768 and lower.  Does the patch you listed enable higher resolutions on the D600?

Posted by Juanito on Jan. 27 2008,06:41
Quote
Does the patch you listed enable higher resolutions on the D600?

- no, it enables higher resolution on the D400 with Intel 855 graphics. Sorry, I was assuming the D600 would have the same graphics as the D400...

Posted by jpeters on Jan. 27 2008,07:18
Quote (Juanito @ Jan. 27 2008,01:41)
Quote
Does the patch you listed enable higher resolutions on the D600?

- no, it enables higher resolution on the D400 with Intel 855 graphics. Sorry, I was assuming the D600 would have the same graphics as the D400...

How about your Xorg72 extension? I'm sure the monitor itself can handle 1280X1024, and is limited by the card.
Posted by Juanito on Jan. 27 2008,07:22
[we're sort of getting off-topic here] - my card works fine with xorg72, give yours a go...
Posted by humpty on Jan. 28 2008,17:26
i'm all for the 'bare-base' approach. how're you going to solve the
'business card' image ? include business.uci with the iso ?

i can also verify that loading/startup and unloading/shutdown of packages
is a lot faster on non-unionfs.

also something i've always been wondering, do these cloop systems
'have' to be compressed? and iso? i.e is there any way to mnt a
tar.gz' (uncompressed)? i'm thinking that tar.gz is more universal than uci.

Posted by roberts on Jan. 28 2008,17:49
I think a bare tiny core may be likely. Details are still being considered.
Much will depend on how much I can easily do with the proto-type.
Other builds with included applications are also an option.

Using cloop could possibly allow to use existing uci extensions.
But of course that could be good or bad.
Bad because of modules and libraries. Not all existing ucis would work.

Does anybody have any experience with archivemount?
This would use fuse to mount a tar.gz.
That could be very interesting if speed is not impacted compare to cloop.

Posted by roadnottaken on Jan. 29 2008,02:04
In the short time I've known dsl, I have gotten to quite agree with your minimalist philosophy.  Unfortunately for me, my computer is a very new laptop that doesn't get along well with the default kernel in dsl 4.x.

This is my request: release the prototype dsl soon, even if it is rough around the edges.  I'm sure that there are many of us who are willing to play with it even if it is somewhat buggy in places.  I know I am.  I just want a chance to have a version of dsl that will run on my laptop.

Posted by lucky13 on Jan. 29 2008,02:12
roadnottaken: While I understand your interest and impatience, I don't think it serves anyone well if it's rushed out. That could be no better than what you (or I, on my laptop) have already. It'll be ready in due time for the usual release candidate process.
Posted by WDef on Feb. 03 2008,01:24
Ping: Robert

re Fuse

A guy I know who who works professionally doing low level linux programming (graphics drivers and such) told me to avoid using fuse because of its overhead.  This is second hand so not terribly high quality.information.

Re: Xorg - might be worth examining how Puppy has stripped Xorg - they seem to have tamed it size-wise a bit..  Having Xorg even in a cut down form would open up acelerated graphics for things like Intel integrated graphics.

Re: modularising dsl - I'd still like to see a typically dsl minimal desktop and little nix apps as the default - maybe have to lose Firefox and do with Dillo or something in order to have 2.6.xx, Xorg and gtk2.  That's ok provided we can load FF as a uci readily.   I  love Emelfm, I do use axyFTP and have used Sylpheed, Xmms, xZGV, Xpdf.  Ted is handy for opening rtfs. Always lamented the loss of Scite since it's my no.1 favorite for programming/scripting.  Don't tend to use the other apps.

I think one reason people don't post so many uncs is they are trivially easy to make with dsl2unc, so if you have a .dsl, you essentially already have an .unc.  I'm not sure the posting frequency necessarily reflects the unc usage patterns.  ucis require a bit more craft.

Only one request though - please continue to compile the loop driver as a loadable module.

Personally, I haven't had any problems with unionfs of which I am aware.  Perhaps it is an immature technology that needs a few years to get worked properly out, at which time you can re-introduce it to dsl.

Posted by jpeters on Feb. 03 2008,04:35
Quote (WDef @ Feb. 02 2008,20:24)
I think one reason people don't post so many uncs is they are trivially easy to make with dsl2unc, so if you have a .dsl, you essentially already have an .unc.  I'm not sure the posting frequency necessarily reflects the unc usage patterns.  ucis require a bit more craft.

Personally, I haven't had any problems with unionfs of which I am aware.  Perhaps it is an immature technology that needs a few years to get worked properly out, at which time you can re-introduce it to dsl.

Exactly....just run dsl2unc and immediately save a bunch of ram.  I've never had any problems either (and I use UNC's whenever possible); I can use a UCI for something that     I'm repeately loading-unloading, which is just about nothing.  Personally, I think it's a technology worth holding on to. As I've once again discovered, what's supposed to work regarding building a UCI (e.g., just configure with --prefix=/opt/) doesn't always work.

Posted by curaga on Feb. 03 2008,07:22
If DSL got gtk2, Xorg, and 2.6, it would get dangerously close to Puppy. What would be the difference then?
Posted by lucky13 on Feb. 03 2008,15:50
Quote
If DSL got gtk2, Xorg, and 2.6, it would get dangerously close to Puppy. What would be the difference then?

I was going to reply in much the same way last night when I read WDef's remark.

My preference would be to try and keep DSL as close to 50MB as possible even if that means moving lots of modules out of the base into MyDSL. I know that size is probably too ambitious with 2.6. The first on my list would be SCSI. Then I would look at other modules on the "extremes" of hardware history -- things that are pretty old and probably less often needed and bleeding edge support. Leave the base usable for "average" computers.

I would leave out X and let users choose to add whichever version they want (a few console apps, maybe add a menu to ease the shock). But I know Robert most likely won't exclude X, so stick with kdrive and GTK1 -- that's enough so anyone who wants "more" can add it via MyDSL. Move things not needed to extend via MyDSL to MyDSL. At a minimum, GTK1 apps like dillo, emelfm, beaver, and sylpheed. Then Robert's murgalua tools. I would remove most of the apps and leave that to user choice.

The reason I oppose GTK2 in the base is because it would increase the calls for including GTK2 apps -- "the latest version of _, so why are you still using that old version?" All of a sudden, you're looking at significant increases in base size because the libraries are already much larger and the apps don't get any smaller between GTK1 and GTK2.

Posted by humpty on Feb. 03 2008,16:22
Quote (curaga @ Feb. 03 2008,10:22)
If DSL got gtk2, Xorg, and 2.6, it would get dangerously close to Puppy. What would be the difference then?

xorg can be separate, the extension has proved that.
puppy runs off root. dsl does not.
dsl has knoppix technology.

dsl could be very modular (depending on the base size).
i was thinking for a 'barest-base' system iso download page, if a choice
of accompanying themed 'basic-apps' could be offered, e.g -
standard, multi-media, gtk-1, gtk-2, diagnostic-disk ..etc in one uci.
this would save the newbie having to select them.

Posted by roberts on Feb. 03 2008,16:28
I will be spending some time face time with John at Scale 6x. This will be a topic that  will be discussed. Basically I agree with Lucky13 on what he posted.

Puppy would not stop me from doing an Xorg/Gtk2/2.6 system. Give me a break. That combination is certainly not unique to Puppy. That is now the standard for most all Linux based distributions.

There are fundamental differences in vision between the two projects. Puppy runs everything as root,i.e., tries to accommodate the novice user. Puppy, I have been told is not really designed to be nomadic, as machine specifics, is by default, in the backup image.

Nevertheless, by adding all of that into the base would mean losing many smaller, less capable systems. Puppy and Austrumi are indeed small distributions but they both use higher level of compression to achieve it and thus both require higher level of computing capabilities.

My view of the direction moving forward is to try to stay under 50MB. Which would necessitate a more "construction kit" approach. Personally I would want 2.6 kernel compiled with gcc 4.2, libstdc++.so.6, gtk1, tinyK, minimal window manger, and all the custom bash/murgaLua scripts that make DSL what it is. Use the MyDSL system for gtk2 and many other applications drawn from the repository. This would allow for the older, less capable machine to continue to move forward with DSL. This is basically what I will be advocating when I am at Scale.



Posted by newby on Feb. 04 2008,01:32
Quote (roberts @ Feb. 03 2008,11:28)
My view of the direction moving forward is to try to stay under 50MB. Which would necessitate a more "construction kit" approach. Personally I would want 2.6 kernel compiled with gcc 4.2, libstdc++.so.6, gtk1, tinyK, minimal window manger, and all the custom bash/murgaLua scripts that make DSL what it is. Use the MyDSL system for gtk2 and many other applications drawn from the repository. This would allow for the older, less capable machine to continue to move forward with DSL. This is basically what I will be advocating when I am at Scale.

I see a few ways to keep most users happy.

1. Build distros with applications geared to different users, office, music, video, games, etc.  Downside: more work.

2. Build application _suites_ geared toward the different audiences and pair them with a base DSL that they can be loaded into.  Downside:  work necessary to maintain multiple suites.

3. Sell a base system along with CD's (or USB keys...) containing the MyDSL applications.  I would happily carry around 2 or 3 CD's or USB keys to have everything I need.  It also gives you a new revenue stream.

Hmmmm.... If one packaged up a suite of MyDSL's inside a super-MyDSL, would everything unpack recursively?  If MyDSL works that way, we users could build and maintain suites of apps!

I'd really like to hear from someone more knowlegable about this last idea.  If MyDSL won't work that way, perhaps it could be modified to do so...

Regards to All,
newby

Posted by WDef on Feb. 04 2008,01:51
Quote
Puppy would not stop me from doing an Xorg/Gtk2/2.6 system. Give me a break.


LOL!  Exactly.  

I think Robert's plan sounds like a wise middle way.   That way we get a light functioning desktop in the dsl tradition, 2.6.xx, a small iso, and an easy path to Xorg etc.

Posted by jpeters on Feb. 04 2008,03:05
Quote (newby @ Feb. 03 2008,20:32)
3. Sell a base system along with CD's (or USB keys...) containing the MyDSL applications.  I would happily carry around 2 or 3 CD's or USB keys to have everything I need.  It also gives you a new revenue stream.

I  just noticed that PCLinuxOS sells access to a high bandwidth repository (server) for a $20 contribution, and confines the CD image to a base system  That might not be a bad way to generate some expense money.

Posted by Juanito on Feb. 04 2008,04:29
Quote
If DSL got gtk2, Xorg, and 2.6...

Well, dsln has 2.6 and gtk2 but not xorg and it's 95MB. I guess dropping back to gtk1 and removing some of the included apps will save some space but then moving from 2.6.12 to 2.6.19(?) will probably take additional space.

Quote
Personally I would want 2.6 kernel compiled...

A late(ish) version of 2.6 would be great for wireless, ntfsmount, video, laptops, etc. Knowing what the kernel & libs of this new version of dsl was compiled from (and having the headers) would certainly be a big help when compiling new apps.

Posted by curaga on Feb. 04 2008,07:25
I did not mean puppy is the only one with that combination. I meant it's the closest competitor. And if DSL made itself 'generic', partly with that move, people would still not be happy, as the libs would need removing apps to keep in 50mb. And the user base would drop.
Posted by clivesay on Feb. 04 2008,14:37
I've been following this thread with interest.

Personally, I would like to see the core as bare as possible with almost everything modular. You can still package some of these modules into a base iso. Most basic users would have no idea that anything is different, however, advanced DSL users would be able to unpack the iso, add/remove modules to create their optimal configuration.

That is my DSL dream.  :)

Chris

Posted by lucky13 on Feb. 04 2008,17:32
newby
Quote
1. Build distros with applications geared to different users, office, music, video, games, etc.  Downside: more work.

You left out larger size. How is it "damn small" if each of your specialty versions is 300-500 MB? Users can use the same modular base in conjunction with MyDSL to accomplish whichever tasks they want, whether specialized or general use. Another problem with creating multiple specialty versions is most people will resort to using MyDSL anyway. Just use the same small, modular base and let users customize it as they see fit.  That's the idea behind modularity.

The downside of it being more work is *the biggest* problem. Robert has had more than enough work maintaining syslinux and isolinux versions of everything, 3.x and 4.x, DSL and DSL-N. Lighten the load, don't increase it.

Quote
2. Build application _suites_ geared toward the different audiences and pair them with a base DSL that they can be loaded into.

Same problem as above. That goes against the Damn Small philosophy, too, by including things that users may not care to have. Your idea of an internet suite would probably be a lot different than mine. Office suite -- ted and siag, Open Office (which version?!), or abiword (gtk1 or 2) and gnumeric? Just let users decide what they want with respect to each extension. Easier -- and ultimately smaller -- that way.
------------
Juanito:
Quote
moving from 2.6.12 to 2.6.19(?) will probably take additional space.

Definitely. And so will moving to a newer kernel if that happens (as it should, imo).

Quote
Knowing what the kernel & libs of this new version of dsl was compiled from (and having the headers) would certainly be a big help when compiling new apps.

I'm sure you'll get all that once it's decided upon.

Posted by curaga on Feb. 04 2008,17:48
When moving to a newer kernel, to take full advantage of it, a new C library would need to be compiled. Current apps would work with it, but supporting them would increase the size of the libc. And this goes back to the point, that to get as fast and small system as possible, complete overhaul would be needed.

Of course, only a new C library would be enough if there's no time..

Posted by jpeters on Feb. 05 2008,00:17
Quote (curaga @ Feb. 04 2008,12:48)
When moving to a newer kernel, to take full advantage of it, a new C library would need to be compiled. Current apps would work with it, but supporting them would increase the size of the libc. And this goes back to the point, that to get as fast and small system as possible, complete overhaul would be needed.

Of course, only a new C library would be enough if there's no time..

DSL already has a great system for older computers with very limited ram. At this point, however, you can get an older computer with a gig of ram fairly cheaply. I recently picked up a Dell Latitude D600 with a gig of ram, rebuilt, for $300.  As an experiment, I loaded Mepis completely toram from disc, and then proceeded to pig out on apps from Synaptic (tcltk, bluefish, etc), in addition to using oo, Thunderbird, and everything else I could think of.  After two days uptime, I still have about 20% free ram; and that's with no swap. No shortcuts on drivers, either; all printers, cards, etc., have plug & played. Also, everything networks by default (just scp to another linux machine, or samba to a windows PC).

Posted by WDef on Feb. 05 2008,20:05
Quote
gtk2

The issue with needing gtk2, unfortunately, is this:  nobody is writing apps using gtk1 anymore.  Most times you go to update a gtk1 app, you find the maintainers have moved to gtk2.

Posted by lucky13 on Feb. 05 2008,22:15
Quote
The issue with needing gtk2, unfortunately, is this:  nobody is writing apps using gtk1 anymore.

But they did write GTK1 apps and continued supporting them before switching to GTK2-only development. There are many GTK1 apps available that are still more than adequate for the purposes of DSL. Just because GTK1 support ceases doesn't mean the old app versions stop working.

I understand your point very well, and I appreciate it. I was very much pro-modularization and came down on the side of sticking with kernel 2.4 and upgrading the libraries when Robert polled us about going forward last year. At the time, I thought it made more sense for the same reasons you think it does now.

If 2.4 development were keeping reasonable pace with 2.6, that would still be my desire. But just as GTK1 development is ebbing, 2.4 development is ebbing even more. Experiencing first hand the frustrations of getting things like ndiswrapper, wireless extension, and wireless tools updated in a 2.4 base, I'm firmly in the 2.6 camp.

Development of 2.4 is nearly at a dead end in terms of new hardware support and many projects are no longer supporting that kernel line. So the bigger issue now isn't in user space with "old" applications, it's supporting hardware. The apps will run on any hardware if DSL can run on that hardware.

I don't think both kernel 2.6 and GTK2 are possible in as small a base as I would like to see, not without removing many modules. Then if you include GTK2 in the base, do you make the base like DSL-N and not support GTK1 at all unless by extension?

Unless it can be done in a tiny core -- hopefully in the same 50 MB frame work we have now -- I prefer to stick with GTK1 in the base and get everyone on the same page with a comprehensive GTK2 development environment so we don't end up with patchwork apps and library packages. That way those who want GTK2 can have it as seamlessly and "damn small" as possible and those who don't want the weight of GTK2 on their smaller system resources aren't burdened by it.

Posted by newby on Feb. 06 2008,16:10
Quote (chaostic @ Jan. 25 2008,02:08)
And as for DSL Future, how about 200~250 mb? That way, it will take up all of a mini-cd, still keeping it small.

Edit:
Oh wait, 512 = ram. I thought size. :/

All of that 250 MB does not have to be the iso image --- I'd put in two isos and use the rest for copies of applications.

One iso would be the current 2.4.x one for older systems.

The other would be a minimalist 2.6.x iso that would load and then load any .dsl, .tar.gz, .uci, .unc extension(s) desired from an "/extentions" directory on the same media where the iso loaded from.

A script could be written that allows the user to _select_ from the /extensions directory.  (Thus getting around the objection, 'your idea of an application suite is not my idea of an application suite.')  Once all selections have been made, a loop in the script would load them one after the other.

A credit card sized boot CD along with a credit card sized applications CD would fit in a wallet.

A round mini-CD still fits in a shirt pocket, two or three of them could still fit in a shirt or coat pocket.  They would still be useful, practical and all that good stuff.  The user base would still be there - no loss to anyone.

AND, the 50-55 MB iso size could still be preserved.

Posted by Nigadoo on Feb. 06 2008,17:37
I'm all for a small base to be extended via MyDSL, as long as the base has enough (X, JWM, etc.) for newbies to get started and easily extend DSL.

Very important to me: giving a new lease on life to old computers with (usually) low ram.  Although I can always go back to older versions of DSL if and when newer versions stop working on the legacy hardware, I would find that a real drag ... it is pretty cool to run the latest version of an actively maintained distro on these old machines!

Here's a newbie-level question: could a base DSL have both kernels available, and offer the user the choice of which kernel to launch, assuming both kernels (and "minimal stuff for newbies") could fit in 50MB?

Posted by lucky13 on Feb. 06 2008,18:20
Quote
Here's a newbie-level question: could a base DSL have both kernels available, and offer the user the choice of which kernel to launch, assuming both kernels (and "minimal stuff for newbies") could fit in 50MB?

No, if DSL stays in the 50MB range it's one or the other.

Posted by clivesay on Feb. 06 2008,18:24
lucky13,

Could that be possible if you stripped out fat apps from the base like firefox? No idea. It is an interesting option. I remember when Mepis and others gave you the kernel boot option.

Chris

Posted by lucky13 on Feb. 06 2008,20:47
Quote
Could that be possible if you stripped out fat apps from the base like firefox? No idea. It is an interesting option. I remember when Mepis and others gave you the kernel boot option.

That choice was one of the things I liked about Mepis-Lite, but the goal with that wasn't a tiny base (just KOffice instead of OpenOffice, etc.).

I suspect any tiny core 2.6 will very likely already not include anything like firefox. My earlier description isn't going to leave very much room unless more modules are removed. You're looking at 2.6 with reasonable modules for "average" computers (meaning vintage and bleeding edge users will need to visit MyDSL to get their modules), tiny X, one very small window manager, Robert's scripts, GTK1, and just enough apps -- dillo, sylpheed, emelfm, axyftp, etc. -- so one can connect to set up what they want. The more modules removed, the more stuff can be included. I think, though, that it's more important to include a healthy variety of modules so DSL can run in the first place. It's one thing to modularize things like apps, it's another to expect new users to be able to understand why the CD doesn't even boot up like old versions or like other distros do without having to download modules separately.

The size differences between kernels are pretty staggering. I don't remember the differences between 2.4.28 and 2.6.12 in Mepis-Lite (iirc -- I'm  more sure about 2.4 version than 2.6 because I only booted 2.4). I have a hunch that both of those maxed out with modules would be close to the same size as 2.6.23.14 by itself.

This is uncompressed:
Code Sample
$ du --si --max-depth=0 /lib/modules/2.6.23.14/
64M     /lib/modules/2.6.23.14/


I admit that's with a lot of stuff that would be moved out as modules because I didn't select "no" to very much. But it's going to be very big by the nature of the beast -- 2.6 is scalable, but scalability isn't easily done in the generic way desktop users need a kernel to be (especially on a live CD intended for use on a wide variety of hardware). FWIW, I'm using the 2.6.23.1 i-686 kernel from Debian Sid on this computer now and it's nine MB smaller than the one I have. For comparison, here's DSL-current on a hard drive install on this drive:
Code Sample
$ du --si --max-depth=0 /mnt/hda2/lib/modules/2.4.31
21M     /mnt/hda2/lib/modules/2.4.31


IIRC, Robert said his tinycore proof of concept was 39 MB and with kernel 2.6.19. Maybe that was earlier in this thread? That gives a *little* room if you move SCSI and other modules out of 2.4.

Maybe it can be done relatively painlessly, but I don't see it as being very desirable for most users. Unless users don't care about 50MB.

That gets to my biggest complaint about including two kernels: it flies in the face of the "damn small" philosophy and you're most likely to only run one kernel at a time anyway.

Posted by clivesay on Feb. 07 2008,03:22
Thought it was worth asking. You made some very good points!

Thanks

Chris

Posted by Nigadoo on Feb. 07 2008,15:18
Although working from very limited know-how, I suspected that 2 kernels within a 50MB limit would be a pretty bit pipe dream.

Even if a 2.6 kernel fits just fine within the goal of a 50MB business card CD, can the perceived goal of a "Tiny Adaptable Linux That Runs on Anything" (from "The Official Damn Small Linux Book; yes, I have a copy and I love it!!!) still be achieved with a 2.6 kernel?  (Don't get me wrong: I'm all for progress.  I just like to squeeze all I can out of legacy hardware before it goes to the dump!)

Is it possible to "trim down" a 2.6 kernel to 2.4 size, with anything taken out of 2.6 made available via modules?

Posted by lucky13 on Feb. 07 2008,16:27
Quote
(Don't get me wrong: I'm all for progress.  I just like to squeeze all I can out of legacy hardware before it goes to the dump!)

Please re-read this thread. IIRC, Robert has clearly stated that there is no intention of making DSL less friendly for older hardware. This has been one of the things that differentiates DSL from other distros, and I hope it continues to be an emphasis (because I won't throw away functional hardware, either).

Posted by curaga on Feb. 07 2008,16:32
There are some tricks, such as optimizing the kernel for size and upx-packing the kernel, which do bring the size down, but not to 2.4 levels.
Posted by Nigadoo on Feb. 07 2008,17:32
Quote
...there is no intention of making DSL less friendly for older hardware...


Thanks lucky13, my misunderstanding.  I thought since older kernels don't support the latest hardware (drivers), newer kernels don't necessarily support some older hardware.

Sincere kudos to Robert and all for doing a bang-up job (overall and specifically supporting old hardware).  I'll vote for anything that keeps Robert and others on this project!  I'll happily use older versions of DSL on the old machines when/if newer versions don't work!

Posted by frankseu on Feb. 07 2008,19:01
Hello,

i would love to have a DSL with ntfs-3g support.

Posted by Juanito on Feb. 08 2008,06:30
Quote
...ntfs-3g support.

I've been starting to look at that (and ntfsmount and bcm43xx) to see which 2.6.x kernel is required as a starting point

Edit: I managed to compile ntfsmount and ntfs-3g on dsln (2.6.12) using fuse-2.6.1 - by the looks of things, the earliest kernel for bcm43xx is 2.6.15, but the bcm43 people are saying >=2.6.22 and are inviting patches for 2.6.18-2.6.21. See < here >

Posted by Juanito on Mar. 18 2008,07:24
@Roberts - any update on progress and latest thoughts on the 2.6 version of dsl?
Posted by roberts on Mar. 19 2008,17:41
To give an update on DSL:

I am waiting on John to make some decisions. But also to let everyone know that personally, I have had some bad news with regards to my vision. As many know I have muscular dystrophy (oculopharyngeal). I have been evaluated for possible surgery to try to improve my vision due to extreme ptosis. There are five type of surgery for this condition. Unfortanuately mine to too serve for any of these to be effective. This means further reduction in vision and therefore limited ability to work on DSL. I need John to once again be involved with DSL developement.

Currently, a decision needs to be made on whether to release a 4.2.6 with a few changes, or a 4.3 with gtk1 firefox v2.0 - still 2.4.31. FFv2 is not only much bigger but also requires much more resources while still, because gtk1, of only supporting flash 7. Which to release is my question to John.

John has agreed in principal for a 2.6 tiny core. But doing so, means we need to have an improved searching ability for the required addons (extensions). Therefore, my attempt to provide an SQL interface to the mydsl extensions.

But lately I have been spending more time with doctors than my keyboard.

-- Robert

Posted by lucky13 on Mar. 19 2008,21:30
Really sorry to hear the news about surgery options, Robert.
Posted by WDef on Mar. 20 2008,04:44
Also sorry to hear about that, as I'm sure everybody is.
Posted by jpeters on Mar. 20 2008,05:21
That's bad news; I hope you've checked with a number of specialists on that, since it's rare that two agree with one another.
Posted by Juanito on Mar. 20 2008,05:43
Quote
John has agreed in principal for a 2.6 tiny core. But doing so, means we need to have an improved searching ability for the required addons (extensions). Therefore, my attempt to provide an SQL interface to the mydsl extensions.

- agreed that would be nice, but could we help with some early testing without the SQL interface?

Posted by Jason W on Mar. 20 2008,06:12
Robert,
 I hate to hear of what you are going through, and my prayers are with you.

Posted by curaga on Mar. 20 2008,13:43
I'm sorry to hear that too. Hope you will get better.
Posted by roberts on Mar. 21 2008,19:50
Thanks for all the kind words, thoughts, and prayers. What I have is genetic and currently no treatment available.
I don't let it get me down. I will continue as I can. It just means it will be less than before.

I am looking at other foundations for the next base. Instead of hacking Knoppix or some other distro, I am looking to build up or even script build up. There are several foundations available.

DSL v3.x will no longer have maintenance releases. I cannot continue to increase my workload. As it is, v3.x is maxed out at 50MB.

As I wrote v4.x will continue but on a much slower schedule. I currently have one staged for release.

The SQL thing is not just for searching. It is to also consolidate the web html tables. The html tables will be a by product of having an sqlite database of extensions. Currently the html tables are built via an ugly awk program that I wrote and wish I hadn't.

So bottom line, I am looking into a new base for DSL.

Posted by jaapz on Mar. 21 2008,22:05
Quote (lucky13 @ Mar. 19 2008,16:30)
Really sorry to hear the news about surgery options, Robert.

Yeah, me too. I wish u te most of luck u can get when u have a/the surgery :)
Posted by newby on Mar. 21 2008,22:10
Quote (roberts @ Mar. 21 2008,14:50)
Thanks for all the kind words, thoughts, and prayers. What I have is genetic and currently no treatment available.
I don't let it get me down. I will continue as I can. It just means it will be less than before.

I am looking at other foundations for the next base. Instead of hacking Knoppix or some other distro, I am looking to build up or even script build up. There are several foundations available.
...
The SQL thing is not just for searching. It is to also consolidate the web html tables. The html tables will be a by product of having an sqlite database of extensions. Currently the html tables are built via an ugly awk program that I wrote and wish I hadn't.

Sorry to hear about it.  I know disability is a bitch.  If it were just the condition, it would be OK.  

I'm sure we all will be interested to hear about your build-up from a new base.

Are you looking into using the database abilities to automate the loading of dependancies?  That would be a significant addition, IMHO.

Posted by ke4nt1 on Mar. 22 2008,16:01
That's not the news I was hoping to have read about your eyes,
although sometimes the 'cure' is worse than the symptoms.

Here's something to drool over till we get together again soon.

< http://www.batista.org/paella.html >

73
ke4nt

Posted by clivesay on Mar. 23 2008,17:07
Quote (ke4nt1 @ Mar. 22 2008,10:01)
That's not the news I was hoping to have read about your eyes,
although sometimes the 'cure' is worse than the symptoms.

Here's something to drool over till we get together again soon.

< http://www.batista.org/paella.html >

73
ke4nt

Or drool over this maybe, Robert? :)

< RS weakness >

Posted by roberts on Mar. 24 2008,16:42
Quote (clivesay @ Mar. 23 2008,09:07)
Quote (ke4nt1 @ Mar. 22 2008,10:01)
That's not the news I was hoping to have read about your eyes,
although sometimes the 'cure' is worse than the symptoms.

Here's something to drool over till we get together again soon.

< http://www.batista.org/paella.html >

73
ke4nt

Or drool over this maybe, Robert? :)

< RS weakness >

Kent & Chris, I would truly enjoy a reunion. We had great fun. Paella, Fraps, Pho, and  much shop talk.
Posted by BobH on Mar. 24 2008,18:04
Well, as a person with his own set of creeping medical issues I too wish you the best.  We have all seen advances in so many chronic medical issues that hope should always remain strong.

You are surrounded by a large group of "supporters" some you know (personally), and some "online".  We all escape at one time or another into our jobs (or hobbies) from the realities of life,  I'm using DSL mostly for "fun" and show it off when I can.  Maybe that's why I keep going back to it even though I have OS X and Windows.

I look forward to the next big (small?) release!  Now if only it worked with my X printer and my Y wireless card!  Just kidding.

Best regards,
Bob

Posted by frankseu on April 08 2008,09:59
I am very sorry to hear about the bad news you got.
Sending you al the best wishes from germany.
regards
Frank

Posted by roberts on April 09 2008,17:37
Thanks for the kind words.
When I am able I am now working on the new version.

Posted by roberts on April 13 2008,03:36
An update:
Kernel and modules are built and booting into a tiny core of Xvesa and JWM with murgaLua scripts working.

Still more specialized modules to build and test.
The tiny core will be mostly app-less and built to support the mydsl system.

I have many changes in mind on how best to deploy mydsl v2.0 based on new system design.

Posted by Juanito on April 13 2008,03:51
Quote
The tiny core will be mostly app-less and built to support the mydsl system.

- sounds great, just what I was looking for  :)

Posted by jpeters on April 13 2008,05:48
This should stir things up a bit within the linux community...
Posted by roberts on April 17 2008,01:40
Update: I have now moved my day-to-day machine onto DSL v5.0
Still lots of work and bugs to fix. Yet it boots, runs murgaLua, and the latest stock Firefox v2.0.0.13 + Flash v9 as on-demand loadable modules.

Posted by Juanito on April 17 2008,06:26
Which kernel version did you decide to go with in the end?
Posted by roberts on April 17 2008,21:12
2.6.24.2. I used Finnix 91.1 as a bootstrap for a base toolchain. A stock 2.6.24.2 kernel and kernel patches from SliTaz. This is how I build the -dsl kernel. Currently some scripts of Knoppix/Finnix/SliTaz are in the cauldron as well as all the DSL custom scripts.


Posted by roberts on April 17 2008,21:24
And more...

It, the core, has no gtk stuff.
Yet, I am able to mydsl-load a gtk2.dsl and then run gtk2 Abiword, Gnumeric, and CUPs.

I also made a gtk1.dsl and was able to mydsl-load it and run our current gtk1 emelfm and LinNeighborhood/Samba extensions.

The proof of concept is working.



Posted by WDef on April 17 2008,23:48
That's great. Having the choice of either or both of gtk1 and gtk2 will be nice.

For example, I haven't managed to get my groove on with the gtk2 version of emelfm.  Somehow it just seems too very.  I still love the old gtk1 emelfm.

Posted by clivesay on April 18 2008,01:31
GEEZ 'O PEETE!!!! I can't stop DROOLING...!!!!!!!!  :D

Great job, RS!

Posted by frankseu on April 21 2008,10:40
Just a question to the 5.0 plans :
Are there planes to have ntfs-3g in the standard or will it be left to an extention ?
THX
Frank

Posted by lucky13 on April 21 2008,12:00
Quote
Are there planes to have ntfs-3g in the standard or will it be left to an extention ?

My vote, if it matters, is to leave that kind of thing to extensions. What's the point of modularizing if everyone's going to lobby to have this or that thrown into the base?

Posted by roberts on April 21 2008,17:56
Quote (lucky13 @ April 21 2008,09:00)
Quote
Are there planes to have ntfs-3g in the standard or will it be left to an extention ?

My vote, if it matters, is to leave that kind of thing to extensions. What's the point of modularizing if everyone's going to lobby to have this or that thrown into the base?

It will not be included in the core.

The core will be ASAP. As Small As Possible.
I am not even planning on having icons.
If one wants xtdesk, dfm, or rox, they too will need to be extensions.
If one wants an icon collection, for even JWM, that too will be an extension.

It does boot into JWM and able to run MyDSL GUI selection tools.

I am not planning on having many console based apps, they too, should be extensions.

As I already mentioned, no Gtk in core. I have a gtk1.dsl and have tested some apps with our gtk2 extension (only as a proof of concept).

I am not going to be very involved with the gtk2 stuff. John, and many in the fourms are way more knowledgeable with gtk2 than I.

I am planning on making adjustments to mydsl subsystem, based on the new architecture. I am going to be supporting less extension types. Doing so will  make the system easier to understand and support. I will have a mounted type extension, with self-contained still being the most desireable.

It is likely that one or several flavors, i.e., core + extns, will be offered.
It is too early to say exactly what will be, as I am only focused on core.

I am going to try to hold the line on what is in core.

Also note that DSL v4.x will continue to be developed and supported as a complete desktop as long as their is interest and available apps/mods to be done.



Posted by jaapz on April 21 2008,21:34
Sounds cool, i really like the fact u chose that 2.6.24 kernel, i think the rt2500 wireless chipset drivers are included in that one, and that would mean that i am going to put 5.0 on my newer laptop :D

Anyway, as far as i can read, dsl 5.0 is going to be just a core dsl with X and some small apps? Thats cool :). Would be easier for u to keep dsl in the 50mb limit, and easier for us to install new programs and customize dsl for our daily use :)

Posted by frankseu on April 23 2008,14:59
I also will be happy to have a core, small Base-DSL and the rest as extentions.
Frank

Posted by lucky13 on April 23 2008,15:57
Quote
I also will be happy to have a core, small Base-DSL and the rest as extentions.

I think it will be well-suited for most hardware already using DSL depending how users intend to set it up. That's one of the things about modularizing it as much as possible that makes it accessible to more users, regardless of targets in the 256/512+ MB RAM area. It's still useful below that if users are mindful about what all they stack on it.

I've been tweaking things to reduce resource drain (using another supposedly "light fast" distro) on my laptop, on which I never wanted to run 2.6. I've made a big dent but still have plenty room to go. But it's definitely an improvement over the default install. I'm kind of in a holding pattern for Robert to release before I go much further in reducing the footprint/demands of what I'm currently using (or making the packages available as a low resource slackbuild).

Posted by roberts on April 27 2008,22:29
Update:

Still making progress on the tiny core. Based on observations, both my own, and recent remarks. Some directions that I am working towards...

I want to try to address many of the older computers that DSL has traditionally targeted. The most often reason cited, from those who know frugal yet still perform traditional hard drive installation, is the slowness from running from compressed read-only images. I now find myself experiencing such with the very large Firefox v2.

Yet I am pleased to see that many are contributing UCI type extensions. With the new system, I want to run those uncompressed. The UCI will morph into mountable application directories. Only the download delivery method will be compressed (TBD). I am planning on only supporting two extension types, our traditional .dsl, which is really a gzipped tarball with menu and optional icon. The second will be the mountable application directory.  

Doing this should provide a boost in performance. With the physical size of pendrives ever increasing, and all that unused hard drive disk space, storing and running mountable uncompressed applications seems natural.

Also is the advantage of avoiding the issues with read-only. No more sym linking out to a writable area. Yet still we have the advantage of simple removal of either the .dsl or the application directory.

For those with much ram, use .dsl, and those without use .map (Mountable Application Directory)

Currently I am running a mix of tradtional .dsl and mounted application directories with better results than everything in compressed loops.

I can easily switch 'flavors' by boot options.

Posted by lucky13 on April 27 2008,22:44
Quote
With the new system, I want to run those uncompressed. The UCI will morph into mountable application directories.

Sounds RISC-y to me.

Posted by roberts on April 27 2008,23:13
Quote (lucky13 @ April 27 2008,19:44)
Quote
With the new system, I want to run those uncompressed. The UCI will morph into mountable application directories.

Sounds RISC-y to me.

I will get there somehow someway  :D

Posted by meo on April 27 2008,23:30
Hi!

Sounds very interesting. I hope fluxbox can be used with it to. I'm kind of attached to this wm since the blackbox days with slackware some years ago.

Have fun guys,
meo

Posted by curaga on April 28 2008,13:53
Looking forward to 5.0 alpha :)
Posted by ^thehatsrule^ on April 28 2008,16:25
Quote (roberts @ April 27 2008,22:29)
I want to try to address many of the older computers that DSL has traditionally targeted. The most often reason cited, from those who know frugal yet still perform traditional hard drive installation, is the slowness from running from compressed read-only images. I now find myself experiencing such with the very large Firefox v2. [...]
How will the base livecd be structured?  Will it be compressed?  If so, would it be decompressed to a 'map' when a frugal installation occurs?

Just curious if these parts have been finalized yet.

Posted by curaga on April 28 2008,16:40
I just thought about the sound aspect. I think DSL 5 should consider OSS 4, instead of ALSA.
While it's not in-kernel, it has many advantages:
- no external libs (saves space compared to all the libasound*)
- greater support, works on all unixes and has been around longer
- many claim it has better architecture, for example it has a software mixer while ALSA needs a hardware one to play many sounds at once

Some links:
< http://www.phoronix.com/forums....e+sound >
< http://insanecoding.blogspot.com/2007....ux.html >

Posted by curaga on April 28 2008,16:45
Edit would have lost the two links, so I'll add here. The download link at 4Front's front page takes to the commercial version, which has some NDA-d drivers, but the open source edition is here:
< http://www.4front-tech.com/developer/sources/stable/gpl/ >

Posted by roberts on April 28 2008,23:43
Just to show it is not vaporware, I booted it in Qemu in 96MB.

< Screen Shot >

Posted by roberts on April 28 2008,23:56
Curaga, currently no sound in core. I am interested in OSS as ALSA is quite large. Not sure sound should be in core. Don't need sound if your use of core is to run a webserver. Perhaps I will release an alpha cut and let you make an OSS extension.
Posted by curaga on April 29 2008,14:12
I'll be glad to help any way possible.

About the .map: will they work from a burned cd? So they would still be readonly?

Posted by clivesay on April 29 2008,15:46
Screenshot teaser?!! Shame on you, Robert! :-)

We're so ready for an alpha!

Chris

Posted by roberts on April 29 2008,15:52
Quote (^thehatsrule^ @ April 28 2008,13:25)
Quote (roberts @ April 27 2008,22:29)
I want to try to address many of the older computers that DSL has traditionally targeted. The most often reason cited, from those who know frugal yet still perform traditional hard drive installation, is the slowness from running from compressed read-only images. I now find myself experiencing such with the very large Firefox v2. [...]
How will the base livecd be structured?  Will it be compressed?  If so, would it be decompressed to a 'map' when a frugal installation occurs?

Just curious if these parts have been finalized yet.

Good question.  I am mainly concentrating on making the core as small as possible and work with much of the DSL basics. Many have suggested that a non-desktop is not useful, therefore some iso should be made available with some collection(s) of useful desktop apps. The catch-22 is that the useful desktop apps come from the community. I don't think you would want me to spend the time to recreate many of them. Saying this, I am somewhat divorced from a combined iso. The answer depends on how the user would use tiny core. With large ram systems the 'other apps' could be .dsl. They could be highly compressed maps to be 'installed' (uncompressed and copied) onto writable media. Or they could be a collection of .map to be mounted and used. Note the fact that such maps on CD would mean read-only and the cdrom could not be removed. At the moment I am more interested in having writable apps be they .dsl or .map on supporting media. Although both you and Curaga bring up valid points. Perhaps the combined iso should be left up to the user depending on needs.

Posted by tagori on April 30 2008,08:49
why don't you drop one of the window-managers and replace the 2.4 kernel with an actual 2.6 kernel?
Posted by lucky13 on April 30 2008,11:23
Quote
why don't you drop one of the window-managers and replace the 2.4 kernel with an actual 2.6 kernel?

An "actual" kernel? The 2.4 one is an actual kernel.

Among other reasons...
1. 2.6 won't fit in 50MB without removing more applications.
2. The window managers are quite small, so dropping them both wouldn't make a dent big enough to use 2.6 unless it were pared way down to about the present config. Then there would be all the issues with it still not supporting newer hardware. The issue isn't supporting the same old hardware DSL has always supported, it's about moving forward and supporting newer hardware as well. That requires balancing things so that you get more support even if it's at the initial expense of having fewer applications.
3. Updating to 2.6 would necessitate updating a lot of the underlying toolchain. Otherwise we're looking at duplicating the present problems often encountered when compiling recent software on older libs.

In short, it's not exactly a matter of switching out kernels. It requires some compromises.

Posted by roberts on April 30 2008,15:40
Quote (tagori @ April 30 2008,05:49)
why don't you drop one of the window-managers and replace the 2.4 kernel with an actual 2.6 kernel?

If it were only that simple. We need much newer libs and there is much newer technology for the new version to be based upon.
Posted by tagori on May 01 2008,08:01
hello

i want to use damn small linux on the asus eee-pc.  this combination would be an ideal 'surfstation' (secure and very fast). but the problem is that the wireless lan resp. the webcam does not work. the eee pc uses nearly the same wlan card then a supported card:
< http://damnsmalllinux.org/wiki...._Box.22 >

a custom version of dsl would be very nice! why dont u make a custom version?

regards, alex

Posted by lucky13 on May 01 2008,11:58
Quote
a custom version of dsl would be very nice! why dont u make a custom version?

Why don't you? There are many remastering guides in these forums and across the Internet.

Posted by roberts on May 01 2008,13:37
Much of the eeePC's harware requires a 2.6 kernel. Likely DSL v5 will be better able to support that machine, by compiling the necessary modules. However, the native eeePC software has been tuned for maximum performance to that particular hardware configuration. Using a distribution designed for generic hardware, DSL included, will never give optimum performance as one that is hardware specific.
Posted by tagori on May 01 2008,13:48
do u already work on damn small linux 5.0 ? when do u want to release the new version?

the eepc is a really great little machine!

Posted by curaga on May 01 2008,13:56
He's been upto something for some weeks now, it'll come when it's ready
Posted by clivesay on May 01 2008,18:18
Hey, Robert.

This thread is getting pretty long. Maybe consider a 5.0 thread to start keeping all the discussion and updates in one place?

Can't wait!

Chris

Powered by Ikonboard 3.1.2a
Ikonboard © 2001 Jarvis Entertainment Group, Inc.