[lugm.org] check this weirdness out

Sun Mar 28 07:53:02 UTC 2010

--- On Sat, 27/3/10, selven <pcthegreat at gmail.com> wrote:

From: selven <pcthegreat at gmail.com>
Subject: Re: [lugm.org] check this weirdness out
To: "LUGM Discuss Mailing List" <discuss at lugm.org>
Date: Saturday, 27 March, 2010, 15:35

On Sat, Mar 27, 2010 at 10:32 AM, Loganaden Velvindron
<gnukid1 at yahoo.co.uk> wrote:
>
> The performance gains that some people claim to be able
> to achieve with freebsd is achieved by offloading to the
> hardware as possible.

:) I didn't claim anything ... i just told you to do the benchmark
 and
compare the results.
So.. if you wish to find some obscure excuses for why it lagged a bit
on the benchmark, you are free to do so, and it sure is great things
to know. BUT :) atleast... for now if freebsd is able to cope for
certain of those things that i need for now (in a faster way).. why
should i change?

I see no obscure excuses, if someone disables offloading in 
freebsd, i'd bet that GNU/Linux will probably be faster. Come on,
the average kernel developer is paid by IBM/RedHat and has
4 Gb RAM desktops running with hardware RAID. All freebsd people
have is a remote cluster to tinker with.

> Look at freebsd's em(4) driver, it's written by Intel Corporation
> and simply included in freebsd without any peer review by the supposed
> developers of ``Free'' BSD. For support, you have to complain to
> Intel. How much is this different from Windows and proprietary OSes
> in
 practice ?
>

The question that i ask myself is how much is THIS different from non
proprietary OSes?

Other don't trust buggy hardware offload engine as FreeBSD does ?

As we can see from the man page of FreeBSD :
-------------------
SUPPORT
     For general information and support, go to the Intel support website at:
     http://support.intel.com.

     If an issue is identified with the released source code on the supported
     kernel with a supported adapter, email the specific information related
     to the issue to <freebsdnic at mailbox.intel.com>.
--------------------

>From
 OpenBSD:

-------------------

AUTHORS
     The em driver was written by Intel Corporation.

-------------------

I have been trying to find some information about what driver is used
for "Linux Intel(R) PRO/1000 Gigabit " on linux.. unfortunately.. huh
there was no doc :p (por maybe i should have searched through the
whole lot of results sent by google?).

Personally, i don't believe drivers should be the problem of the OS,
but the hardware manufacturers... they need to sell their drivers so
they've got to do sell a complete product..

And what happens when the hardware is EOLed ? We let the driver
rot or lock the API to support backward compatibility ? Nice way
to get a ``Free'' OS to evolve.

Also, am not arguing saying that FreeBSD is perfect :p it is't, Linux
isn't perfect either :D.
You don't like it, you don't use it.

That's why I'm warning
 people against using it. I dislike FreeBSD in
its _current_ state. However, if they start fixing their code, then I wouldn't
mind recommending it to other. Fair right ?

> The Linux developers have long opposed this practice because it complicate
> matters and makes debugging very difficult. (Is it in my code, or in the firmware of
> the ethernet controller ?).
>
> See this:
> http://www.cs.helsinki.fi/linux/linux-kernel/2003-28/0029.html
>

So you want to impose the Linux way of managing a project on FreeBSD? :)
Communist! :p

Better be communist and get something decent out the doors. As the
saying goes `` A camel is a horse designed by a committee'', and freebsd
has confirmed it again.

> What the Linux people try to do is optimizing their code, while
> the
 freebsd people
 have taken the lazy approach to improve
> performance.

Which is why with a simple benchmarking with mails and AVs they've
lagged behind a crappy OS like FBSD?

Read this :

http://www.cs.helsinki.fi/linux/linux-kernel/2003-28/0029.html

> As for the freebsd docs, it's good :-) It's the same as saying
> ``Microsoft has MSDN, FreeBSD has something similar too !''

True indeed, the only difference is that the handbook is more readable
that MSDN.

Thanks to you I believe we should all stop using OSes that have
documentations... Let's go to the source each time we need to find out
the API of something. :) Now i can see how fast we can concentrate on
problem solving more rather than fight with the machine :)

The handbook has a very corporate feeling, glad you like it.

>
 I'd largely
 prefer to trust Linux for running a large volume mail
> server, at least I know the developers can fix problems if my
> ethernet controller hangs up (It happened for a production server
> at dedibox), instead of trying to get 5% performance gain and limited
> support :-)

:D Is it faster to get the problem solved by waiting for the
developers to fix it or to throw in $1000 and get a stable and
supported Ethernet card?

$1000 for an ethernet card ? We paid $100 dollars monthly/server.
Are you putting gold plates inside that ethernet controller ? :-) 

I guess you'd say wait 2 months for the developers to fix the problem.
Sure if you want, you can.. suits you :) ... your competitors will
wait because they can understand a small driver issue i guess...
praise be the good hearted competitors out there ... amen.

took around 1 week to fix the issue. Halting activity for one week
to get a
 stable ethernet driver, rather than riding months with spotty
driver issues is called what I'd called long-term planning :-D

> As I said previously, save yourself the trouble and use GNU/Linux !

Well, I'll say... Save your soul and use the Open Source project that
suits your needs best instead of running after perfection :)

It's not about perfection, it's about being far from crappiness as possible,
and getting least number of negative feedback. FreeBSD has a number
of issues currently. If they get their act together, I wouldn't mind 
re-evaluating it ;-)

ps. OS wars were never productive, its an open source project, you
find something that needs to be changed.. join in and contribute :p

I'm a user.

pps. The Only Perfect OS i know of is the GNU/Hurd ... we all know its fate :)

GNU/Hurd is broken. It's mostly about changing micro-kernels with
what is currently the buzz-name
 in academic circles. 

^^
+selven

__________________________________________________________
Linux User Group of Mauritius (LUGM) Discuss mailing list
Website: http://lugm.org
Mailing list archive: http://lugm.org/pipermail/discuss_lugm.org/
Forum: http://lugm.org/forum/
IRC: #linux.mu on Freenode

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://discuss.lugm.org/pipermail/discuss_discuss.lugm.org/attachments/20100328/35772be4/attachment.html>