Linux and the desktop (ot)
Posted Nov 1, 2002 12:15 UTC (Fri) by pichwo
Parent article: Linux and the desktop
sorry to be ot, but i am missing the "linux on the server" topic.
i think it is a clear design to build a operating system from bottom up with including it's full desired functionality in eyesight.
desktop users want to have a fast and "no-know"/quick-and-fast-setup system running. they do not care about the things too much. it has to work immediately and never touch a running system (or experiment it to death).
servers need clear blocks of long-term, well designed os support.
desktop hype (OH I REMEMBER the M$ race in the nineties too well) obviously wants rapidly introduced new features. kind of new-to-click-now-junkies.
it is legitimate and ideal for osdn to go with this market, but the overall development branches for LINUX as a base system must be oriented on the blocks of kernelcode, which is needed on typical servers.
let the desktop occupy the rest of the kernel, but i think there should be clear priorities.
this code should have a clear desingn, interoperate well and SHARE functionality, not reimplement things for the N-th time.
e.g. storage managment (this is where actual lvm-dilemma, hooks in)
xor/mirr/striping physics should be handled by a clear md managment, since it is far too close to concrete io-paths.
one should clearly know, what one does when organizing md's.
these decisions needs a clear level-0 managment (software & brains).
how to rearrange/group/peruse those chunks of storage based on their abstract characteristics is a clear level-1 decision.
so why introduce evms ?
quote some sysadmin (even on lwn), who said "i want to have this all from one gui" ?
e.g. filesystem managment
no big deal to install any *fs and to obeye or give a damn on the subtle interdependencies with the underlying physics and high-level io-characteristics.
but how about snapshotting, cow, ... call it what you want.
it is all about the same : transaction oriented (delayed, assured, dropped, whatever ...)
(i admit, it is not a desktop-issue :-)
where is the clever set of kernel-modules which will handle this alltogether separated from the **-trees in WHATEVERfs ?
as i said, facts like these might seem ot here.
i like the diversity in linux very much, since it forces competition and yields an overall gain in quality.
but it is my strong credo, that the battle against other os'es will be decided by parameters like
- clear concepts
- clear abstractions
- clever modularity
and please keep in mind, that the #1 desktop is successful because it is a bastard.
other people might say : ... despite it is , and are wondering
or even worse : dont care about the system, the end-user impression must be customer-oriented and easy-to-use.
i object to ruin a potentially good operating system like linux by trying to compete with bastards by getting a bastard too.
and i would not write these lines, if i would not have the feeling, that some developers are drifting towards this direction.
to post comments)