On 2/9/07, Norbert P. Copones <[EMAIL PROTECTED]> wrote:
On Wed, February 7, 2007 9:16 pm, Dean Michael Berris wrote:
> I had been developing a C++ application back then which used Berkeley
> DB and had been developing with the target platform being FreeBSD 5. I
> had to check why I was getting extreme corruption rates on the BDB
> databases, and low and behold it was a UFS issue -- because it never
> really does a "sync()" on its own, and running on an SMP machine
> caused race conditions and data corruption in high read+write systems.
>
> Now that was already FreeBSD 5, and the touted stability crumbled
> right before my very eyes. And no, it's not a trivial system that I
> was developing. But you know what? The application ran better on an
> un-updated RedHat 9 system which featured a 2.4 kernel using the then
> "new" ext3 filesystem with journaling. No, I didn't have to do
> anything with the code because I was using the same compiler in
> FreeBSD and Linux, and I had been able to do a side-by-side
> comparison.
>
> So the last time I checked was FreeBSD 5 versus RedHat 9. Go Figure.
hmm freebsd5 was never considered stable even by the developers. the smp
code is new, the fs code is also new at that time (ufs2). freebsd5 branch
is considered a "TESTING" branch. its an experimental branch leading to 6.
you should have used freebsd4 on that time is you want stability. not
starting another flame thread but you should have read or asked a local
bsd group why that thing happened on you instead of concluding that one
os' fs sucks.
Wait, so you're saying even the developers consider FreeBSD 5 unstable
-- but why was it released and given a "stable version release" number
if that was the case?
And the fundamental design flaw of UFS -- the latent syncing -- has
been a big problem for the longest time on UNIX systems. That's the
reason why SGI came out with their own filesystem and so did IBM and
other UNIX stakeholders.
The conclusion here is UFS sucks because it by design is flawed.
Nothing more, nothing less. And since the BSD kernel still uses it by
default shows how much better Linux has become as far as being
technically superior is concerned.
As for the SMP code being new, that shows how slow the development is
in the BSD kernel trees and how way ahead Linux kernel development is
as far as symmetric multi-processing is concerned.
you can retry the test using either freebsd4 or freebsd6.
Why should I bother, when I can get the latest and greatest
technically superior solutions in Linux? And when I can be assured of
high performance and stability that no other OS can provide, why
should I even bother with BSD when obviously Linux is technically
superior?
Remember, my comparison was with a very ugly Linux distribution (RH9)
which contained an old Linux 2.4 kernel and even that beat FreeBSD 5.
> I don't F'n care whether the US government funded this or that -- that
> doesn't matter. Linus didn't need a grant to develop Linux and it's
> even better than any BSD kernel in any aspect IMNSHO.
in any aspect? does linux have carp? does linux do pfsync? does linux have
pf?
What are these and why would I need it? When in Linux, what I care
about is that I get the performance I bargained for and be assured
that the solutions are of the utmost quality. Old code does not mean
better code -- and I will always settle for better code than anything
else.
yes it does have iptables. but have you considered comparing the two?
i have also read a thread in the linux kernel mailing list that linux is
lagging behind freebsd in terms of wireless drivers. iirc andrew morton is
even worried before about the status of the linux kernel development.
linux may be better in "some" aspects but bsd is also better in "some"
aspects. if you say in ANY aspect, we can't agree on that matter but that
doesn't mean another thread war should be started again.
If you actually care to check, Linux kernel development has been
better than ever. 2.6.20 is coming out soon, and packs more than any
other OS kernel can even pronounce.
If it's beginning to sound like I'm a Linux Fanboy, then yes I am one.
Not because I haven't tried others, but it's because when you start
reading good code going into something as important as an OS kernel,
you appreciate the effort and the technical superiority of the
solutions you're using. And once you've got good solutions to common
problems, than you got my vote. Otherwise, using FreeBSD or NetBSD or
OpenBSD for nostalgic purposes and for feeling a bit of "connection"
even if it means sacrificing performance to keep old code in the
kernel doesn't appeal to me very well.
That being said, on technical merits Linux will always IMNSHO beat any
BSD kernel in aspects that matter: performance, stability, and
efficiency.
If you want to argue, show me code and let's see which technical
solution is better than the other. Then tell me again why BSD is
better than Linux.
--
Dean Michael C. Berris
http://cplusplus-soup.blogspot.com/
mikhailberis AT gmail DOT com
+63 928 7291459
_________________________________________________
Philippine Linux Users' Group (PLUG) Mailing List
[email protected] (#PLUG @ irc.free.net.ph)
Read the Guidelines: http://linux.org.ph/lists
Searchable Archives: http://archives.free.net.ph
