Why NetBSD? (Score:5, Informative)

I believe NetBSD 5.0 is a major turn of tide. Compared to 4.0, this is definitely a new chapter. In a way Mr. Hannum did a favour in his infamous rant: practically all aspects he identified have been addressed.

We here at $DAYJOB have made extensive evaluation of the NetBSD 5.0 pre-releases and it is looking very good indeed. Our internal benchmarks show that for our typical workload, performance of NetBSD is now comparable to that of Linux and FreeBSD. (Numbers and methodology may not be representative nor even correct, but we have to base our decisions to something.) It is very likely that we will be rolling the next big-iron production line solely with NetBSD again. The recent happenings with Sun and the uncertainty surrounding Solaris have warmed also the management section upstairs.

Besides performance and SMP, other things that account high in our book:

1. Long support cycles and backward compatibility. This has always been one of the greatest strengths of NetBSD.
2. Stability. When evaluating an operating system for servers that should be online to the late next decade, it is of crucial importance that stability is guaranteed. In this item there is a question mark for Linux in our list.
3. Security. We believe that reducing the amount of code running is the first step to a more secure environment. In the Linux world the trend has been exactly the opposite even with so-called enterprise distributions. From NetBSD 5.0 we look forward to ASLR and security features inspired by PaX, the technology which we believe in with our Linux systems. This is one area where there is a big minus sign for FreeBSD.
4. Xen. Not only does it run on NetBSD, but does it better (stability-wise) with NetBSD. Contradicts with all previous points, but is probably going to be employed in some parts of our farm, even if it is just to please management.
5. Journaling support. While ZFS is the clear winner here, no doubt about it, the so-called WAPBL has looked promising in our tests. Roughly speaking, our conclusion has been that ext3 and FFS+WAPBL are quite comparable in their performance and stability. To rephrase this: both suck equally bad. Our Solaris-fileservers are not going anywhere for at least five years.
6. Cleanroom implementations and central source for code. Since we have a relatively large in-house software stack, this, combined with (1) and (2), is very important to us. We are an open-minded shop with technically competent people who can replace non-optimal or bad parts of the system with internal designs. Avery dangerous thing to do with Linux due lack of coherency, but our previous experiences have shown that this is not so much of a problem with NetBSD.
7. Documentation. Add to the end of (6). A big minus sign for Linux.

Some drawbacks:

1. The package system. A huddle of shell scripts without a strict API. Can be forgiven since nothing essential depends on it. A big plus sign for Linux.
2. Lack of binary updates. A blessing and a curse. Can be forgiven in our case, but would be a pain in more heterogeneous environment. Again a plus sign and a parrot sticker for Linux.
3. Java. Practically a complete lack of enterprise-like support for this nasty piece of software rules out NetBSD in many of our servers. Linux and Solaris are the only options to consider.

At $HOME perhaps the most exciting feature is the new power management framework. This has taken huge leap forward in NetBSD 5.0. While there is still much work to be done, the direction is right. I believe that like SMP on the other end, power management will be one of the dominant factors in consumer-grade computing at the other end of the spectrum.

Other things that I like generally in NetBSD:

1. UNIX legacy. I have always liked the history of computing and it is a fresh breeze of air to find a system in which you can still feel the good-old days instead of the GUI of today.
2. Comprehension. This is very important: I can actually understand how the system works. Please do not take this as a flamebait, but with each release of modern desktop Linux distribution, I find it harder and harder to actually understand what is going on under the hood. And I am a system programmer by education. (Quite possibly this is just the price to pay from entering the consumer mass markets.)
3. Community. This is also very important to people like me who just like to "get on with it". 100 % technical discussion guaranteed. "No nonsense" could well be the alternative motto of NetBSD.

I could go on with the list, but what a heck, BSD is dying and old farts like me deserve to die in the brave new world of computing...