Re: [Cooker] Re: [Contrib-Rpm] kernel-multimedia-2.4.21.1mm.6mdk-1-1mdk

[eddie]

Buchan Milne wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
FACORAT Fabrice wrote:

Le mar 12/08/2003 � 14:26, Buchan Milne a �crit :


FACORAT Fabrice wrote:


I just wonder if kernel-multimedia should not be included by default
when the user make a desktop install ( ok, there is not really a
desktop/server class during install, so maybe propose if the user want
to optimise is system for multimedia and if Y, then install
kernel-multimedia instead of standard kernel ).
mm kernel is in contrib => not on CDs.


so it may be on CD.



Must go into main first.


Here are my points :
        - normal user need a working supermount ( no erratic CDROM/floppy
access pb, no pb when transfering a large file from CDROM to HD ). Most
of the time we advise mdk to newbies because they will have a simple
installation, good config tools, and ... supermount. Indeed newcomers
don't appreciate and understand well mount/umount mecanism and they are
fed up with it when they try to change the CD and they can't because of
a console or even fam after having access it wia konqueror. So
supermount-ng is interesting for them.
Supermount should work in the stock kernel, or should be removed.
MCSE-style server administrators also want reliable access to CDs.
supermount is not a necessity,


No, it's a convenience, which newbies should enjoy.


and MSCE-style server administrators are
bad IMHO.


They have to start somewhere, and they are a big market. Anyway, RH has
automount enabled by defualt on their GPL edition.

But mp3 support has nothing to do with kernel patches really.


yeah, it as to do with putting bledding edge features, and what desktop
users wants.


Mp3 isn't a bleeding edge feature I think.


RH target enterprise ( workstation/server ) and most of the time they
don't need mp3 and in order to not have pb with patents ( USA lawyers
are sometimes crazy ) they don't put mp3.


Mandrakesoft approached Thompson, but RH claims their lack of support
for MP3 is a software patent protest.

But desktop users complains
about that.


standard kernel should be for server like activities : best stability,
throughput, acl, security, scalability. You don't need supermount on a
server, or low latency.
[*] Note : kernel enterprise is suitable is you have more than 1Gb of
RAM. if your server have less, you end up with normal kernel, so
kernel-enterprise is for server with ... many RAM.
Now if with HIGHMEM support a kernel is not slower or too slower than a
kernel without it, standard kernel could be remove and so u have just
kernel-enterprise, kernel-smp and kernel-secure.
What about low-end servers?? We have a P133/32MB firewall (currently
MNF8.2), I don't want to lose some ram to HIGHMEM support.


so that's why I ask if there are issues with HIGHMEM support. So you
point one ( mem consumption ). others one ? Now you can even balances
advantages/issues in order to see what to do. Does mdksoft want to
target low-end server, etc ...


When we could be stealing market share from RH AS 2.1 or SLES 8, the
only distro's certified for Oracle clusters? I don't think we can afford
to not target low-end.

Hmmm, should be more stable than the competition, which is not trivial
these days.
at least more stable than windows. stable enough means to me that if you
can't have an uptime of 2 years with a daily normal use, it's not a
problem.


So I understand if Juan Quintela don't want to include supermount-ng,
Instead, include a supermount with known bugs?
It's what mdksoft do and seems to keep doing ...


And you think it's better having a desktop kernel available than to have
more bugs fixed in the default kernel? Should we not have the updated
from Thomas fixing server drivers (is AIC7xxx in stock kernel up-to-date
now?).


low-latency and maybe other fancy stuff. But for desktop user ... it's
good for server, maybe workstation, but not for desktop user. Most of
the time they want bleeding edge. Ok, mdk provide kernel-multimedia on
third CD, and ? Are you going to tell the user to install by hand this
kernel after the install ? and if they are using for example nvidia
drivers ? they install kernel-multimedia, reboot, black screen, "help,
help ... blabla - you need to reinstall nvidia drivers - how can i do it
?
Well where did they get NVidia drivers for their kernel?


http://www.nvidia.com
and most of them will go there, and most people will advise them to go
there, and as windows users they will tend to go there.


Well, then they know how to fix it, they can't complain "help, black
screen".

We could
consider making NVidia driver RPMS for mm kernel for Club for example.
maybe, but most of them won't be in mdkclub.



Maybe more will be in MandrakeClub when they realise it provides things
like this (ie being able to setup Nvidia via XFdrake after entering Club
memberhip details should work, not command-line use at all).

IMHO, Juan needs to apply the patches sent to him, otherwise the stock
kernel has problems in either role.
providing one kernel for desktop/workstation/server is bad idea as aims
are not the same.
tmp/mm kernel are usefull for desktop users, but not for enterprise
users. 0(1) as corner case pbs, the same for low-latency and/or preempt
patches, etc ...

Mandrakesoft may not want to officially support the -mm kernel (means
providing/testing kernel updates for it also).
That's the pb. mdksoft target desktop/server/workstation market. But
when you look at the distro it's like a batard, it's not natural.
i.e u can use for desktop, but it miss some things.


Like what? Does any other disto ship a low-loat, preempt kernel?


You can use it for
server but all this bleeding edge packages make u feel that you may end
up with something not sure.


So kde-3.1.3 on your server is bad, but it's ok to have samba-2.2.3a
(Debain stable, ancient release the samba team won't support anymore) or
samba-3.0.0beta2 (Debian unstable), or maybe just samba-3.0.0beta3 (RH),
when we give the best samba-2.2.8a packages (except maybe for SuSE)
*and* the choice to run samba3??
Sorry, I don't buy it. Also, Redhat's glibc in RH9 was so broken you
could not use it in any environment with large numbers of users, you had
to wait about 6 weeks after release to be able to use it in a large
Windows domain (for example).

mdksoft at this time seems to not have the ability to cope with this (
they need maybe more people and of course more more money ).
mdk have choice to do.
they want desktop ? so they should do more like Lindows/Xandros.


It's not the look that needs to be addressed, it's the management of
thousand's of desktops (ie NDS/AD for linux), and storing msec config
and urpmi config in ldap could be a good way to start ... (ie ability to
automate/delegate sofware installation, and ability to lock-down
security en-masse).

Look
what happen with Wal-Mart. Lindows is best fit for desktop ( customised,
windows like, less apps leaving for clarity ).


But Lindows makes no money here. The market is corporate desktops, not
home users.

they want workstation/server ? so they should do as Suse ( special
workstation edition, fully customise ) or RH.


Too expensive, and you need 3rd-party apps certified. Best market IMHO
is corporate uniform desktops (ie not scientific-specific use, but
office-work).

They want want firewall market ? do a special version for firewall. They
do it well with SNF/MNF. The product is customise and provide config
tools customised for it.


BTW, snf packages work fine on 9.1, I have about 6 MNF boxes in
production, only one is really MNF8.2 ...

They need to see what they want to target. When they know they should
customised the distro for them and thus by providing several specific
distro.
With their swiss knife you will have people using it but IT and big
vendors not. Look the HP deal. They provide mdk on their workstation but
not on their server, they'd rather use Suse ( or RH ). Why ? notoriety,
specific products customised for this task.


Well, IMHO, Mandrake is better on a server than Redhat ... and saves
admin time ... the problem is that Mandrakesoft doesn't sell the
server-side, probably too expensive to do so ... but Mandrakesoft needs
a hardware partner for server-side, pity about HP, Mandrake does quite
well on Compaq servers ...
Regards,
Buchan
- --
|--------------Another happy Mandrake Club member--------------|
Buchan Milne                Mechanical Engineer, Network Manager
Cellphone * Work            +27 82 472 2231 * +27 21 8828820x202
Stellenbosch Automotive Engineering         http://www.cae.co.za
GPG Key                   http://ranger.dnsalias.com/bgmilne.asc
1024D/60D204A7 2919 E232 5610 A038 87B1 72D6 AC92 BA50 60D2 04A7
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQE/ORV2rJK6UGDSBKcRAlkqAJ9zCZusl7XKX6H23ohlkHbhbQBfeACdGuQE
l2HEaiaYnXFRmA3yQVxc20o=
=Xj7i
-----END PGP SIGNATURE-----
******************************************************************
Please click on http://www.cae.co.za/disclaimer.htm to read our
e-mail disclaimer or send an e-mail to [EMAIL PROTECTED] for a copy.
******************************************************************

Sometime it's this type of thinking that holds Mandrakesoft back, in my 
opinion. If the mm kernel is better (working without a hitch) why 
wouldn't MDK use it? So what that the kernel team didn't come up with 
it, if it is a better  working kernel, USE IT! The same goes for Thomas' 
kernel, it seems to be going very well. If MDK won't/can't use these 
improvements I think their vision is rather narrow. I have been under 
the impression that if "Joe" can build a better "mousetrap" at least 
take a look at it without condemning it to "contrib". What is this 
monolithic "contrib" and "main"? If someone could enlighten me as to 
what rules make this up, I would glady listen. My 2 cents Edward A 
Mihalow Jr-Silver Club Member