FreeBSD vs. Linux vs. Windows 2000 comparison

[Mark J. Dulcey]

David Kramer wrote:
> I stumbled on this page today.
> http://people.freebsd.org/~murray/bsd_flier.html
> 
> Do you feel that this comparison is still accurate?

We must understand that the comparison was written by a FreeBSD 
partisan. Naturally, there is a certain amount of bias in his assumptions.

One of the downticks that Linux receives are because of its default use 
of a non-journalled write-behind file system. It is indeed true that 
this may not be the best possible choice for a mission critical server. 
On the other hand, it probably contributes to the excellent performance 
of Linux on modest legacy hardware; desktop users, especially those 
using older computers, may be willing to trade off a bit of data 
security for better performance. Recent distributions make it easy to 
choose a different file system; ext3 on Red Hat, reiserfs on SuSE. jfs 
is now in the source tree for 2.5; we'll see future distributions that 
make it easy to use it.

The author also objects to the very open development process used for 
Linux, preferring instead the less widely spread development model of 
FreeBSD. There are indeed strengths to the FreeBSD model, but there are 
also advantages of the Linux model that the author does not address.

He objects to the splintered nature of Linux. But that's sort of an 
apples-to-oranges comparison. One could argue that the relevant 
comparison is to the entire BSD world (FreeBSD, NetBSD, OpenBSD, BSDI, 
etc.) - and across that field, you see the same sort of compatibility 
problems that you sometimes see across distributions of Linux.

He also fails to cite some of the strengths of Linux - and some of those 
come from the scattered development model that the author decries. For 
instance, he says Linux is weak because of the fact that its developers 
discourage the distribution of binary-only device drivers. But he fails 
to note that Linux actually has drivers for more devices than FreeBSD 
does (many coming from those scattered developers) and that the social 
pressure for open source has indeed caused some companies that 
originally withheld information from the open-source community to change 
their policies.

He also doesn't talk about ease of installation and administration. 
Modern distributions of Linux come with good GUI tools for those tasks. 
On FreeBSD, you're still looking at working in character mode. Working 
in text has its advantages, but I'd still give some advantage to Linux 
for giving you a choice.

Finally, I'd give Windows a neutral rating at best on device drivers. 
The problem with Windows driver support is that Microsoft doesn't exert 
any pressure on developers to provide support for older devices on new 
versions of Windows. It's not uncommon to have to retire 
still-functional hardware when you upgrade. Linux, by contrast, has an 
excellent record of maintaining support for legacy hardware, though I'm 
disheartened by the decision of the XFree86 team to ignore a lot of 
older video hardware in their 4.x releases.

And to correct his statements about the Windows file systems, NTFS was 
indeed designed for multi-user systems. It's actually a reasonable 
design, comparable in technical merit to the non-journalling Unix file 
systems; it borrowed a lot of ideas (and some of its development team) 
from VMS. It does have to do the icky stuff with short and long versions 
of file names so that the OS can provide compatibility with old DOS 
software, but I don't think it's as bad as the author says. (But I'd be 
a lot happier if Microsoft would release full information about its 
layout, so that we could have reliable NTFS partition support for Linux.)

Quoted from the page you cited:

> Since both operating systems are open source, beneficial technologies are shared and for this reason the performance of Linux and FreeBSD is rapidly converging.

I couldn't have said that better myself. The open nature of these 
operating systems works to the benefit of both, and to the benefit of 
all their users.

FreeBSD is indeed a "tighter" system than Linux, probably because it is 
developed by a small team. Everything is neatly packaged and works well 
together. I'd consider it strongly for server use. The greater diversity 
of Linux, on the other hand, certainly makes it a lot more fun for the 
desktop hacker; there's a lot more stuff to play with, and there's a 
better chance that a driver exists for that stray piece of hardware you 
have lying around or pick up at the MIT Flea.