Slashdot Mirror
Is BSD Dying?
Every BSD article posted, trolls come out and post about BSD dying. Naysayers at every turn, mostly pro-Linux, say that BSD doesn't have the marketing and advocacy to succeed. Greg Lehey, author of The Complete FreeBSD and FreeBSD core team member, takes a look at naysayer's claims, the history of BSD, the root of the "quiet" BSD advocates, and the relationship of Linux to it all, in this month's Daemon's Advocate at Daemon News
Does BSD have anything similar to apt? Not ports.
/usr/ports/net/bind;make install, to get the new bind. Right?
Ummm, why not? It automatically pulls in the source code, applies FreeBSD specific patches to it, compiles what you downloaded as well as any other dependencies it may need.
Ports (AFAIK) are only for installing packages, not upgrading them.
Yes and no. At present there is no "make upgrade" kind of command for the ports tree. A user must first either perform a "make deinstall" or "pkg_delete" to remove the old app. At this point, all you need do is install the new version.
On an additional note, there's been a lot of talk about working in an upgrade ability to the ports tree on the FreeBSD mailing lists. I've only briefly been following the conversation, but it seems that a couple of folks have made some serious progress in this regard. It seems that much of the problems with this are with the process rather than a technical difficulty. The devil is in the details sort of thing.
For example... Say I install bind through ports. Then cvsup my ports, and the new ports contain an upgraded bind. I need to cd
Your example actually has a couple of oddities to it that you probably aren't aware of, but you've essentially got the idea. I don't see this as a weakness of the port system, nor as a strength of Debian's manner of doing things. By building this into a directory structure it allows a user to sort of browse through the apps they might be interested in. If anything, it may be time for a wider variety of port categories. For example, I have trouble thinking of StarOffice as simply an editor.
As to the oddity of your example, I believe you'd have to run a make world on the entire OS for the latest Bind updates, as they are a part of the core system. Perhaps a more experienced BSD'er can help clarify this point for both of us here. Your example is still relatively valid, as this would be the process for pretty much any other port.
If I ever get back to trying Linux again, it'll most likely be Debian. Darn near installed it to play around a while back. The installation instructions just zoomed over my head at the time, where as FreeBSD was 2 floppies and a handful of menus. Today, FreeBSD is doing pretty much everything I wanted a Unix based box for, so I doubt I'll have a reason to give Debian another try any time soon.
As you said, each to their own.
The line must be drawn here. This far. No further.
BSD is not dying. I am both a Linux and a FreeBSD user, and I can assure you that its not dying. Those anti-BSD "trolls" are nothing more than ignorant Windows lusers that have carried over their misinformed ideas from anti-apple/mac to anti-BSD. As Linux gets easier to use, there will be an ever growing stream of these morons.
Just go to netcraft.net and take a look at the top 50 uptimes. Only 3 times does any non-BSD OS appear there (BTW, Linux does not appear once). BSD is not dying, Linux has just stolen the spotlight for a few moments. Sure Linux has the spotlight, but BSD remains running on all of the mission critical servers. It doesnt matter how popular Linux gets, there will always be a place for BSD; this marketshare is not really being stolen from BSD, its being stolen from Windoze and others.
It doesnt matter how many millions of dollars some useless company is worth, the OS will still be pretty much the same. You cant just pump money into an OS and expect it to be better. BSD is a good product because of its age, and the truley devoted programmers. 1 free programmer doing what they truley love is the equivelant of 10 hired/commerical programmers. Quality, not quantity.
There are some things that Linux is good at, and some that BSD is good at. No amount of advocacy will change it. Advocacy is not the answer in most cases, it just turns into a flame fest. Once people see the statistics they will know in their heart.
Buying a Dell computer is equivalent to dropping the soap in a prison shower.
You're assuming that FreeBSD will stand still while Linux 2.6 is in development.
And you also make the ridiculous assertion that there is some sort of absolute performance metric, that linux 2.4 has narrowed. There is no such thing as "better performance" there is only better performance for your workload. For some workloads, even Linux 2.2 is better. For others, FreeBSD is wildly superior.
There is no magic "go fast" register that FreeBSD pokes. Its optimized quite well for some tasks, and Linux makes slightly different choices.
For example, Linux's poll and select implementations are signfigantly faster than FreeBSD's for small numbers of FDs, while FreeBSD's is faster for large numbers. FreeBSD pays the constant up front overhead to lower the per FD overhead. This means that wrt to poll, Linux is better turned for running X clients, which are just loops around poll() with 1-5 FDs. FreeBSD on the other hand is more optimal for the case of a huge number of FDs, as found in large fileservers.
There is no magic algorithm that will make both cases run optimally. You have to choose one or the other. Another tradeoff is latency vs. throughput. Memory usage vs performance (usually caching or precalculation).
OS design is an exercise in tradeoffs. There is no one perfect OS, nor can there be. If people are advocating one OS the best (be it linux, solaris or freebsd), question their motives. Perhaps preserving job security and unwillingness to learn new things is the real root of these flamewars. I know I've seen that in DBAs who know only Oracle, and try to tell me its the best for multidimensional OLAP workloads.
As someone who has worked with Linux, NT4, Win2K and FreeBSD in (yes) mission critical datacenter environments, I would say that FreeBSD is my pick (Solaris being ruled out on cost).
Lately though, we have been forced to adopt Linux simply because the latest Oracle & Java 1.3 environments are unavailable on FreeBSD. At the end of the day, an OS is as good as the applications it supports. DNS, qmail, Apache rocks on FreeBSD. But where are Oracle and Java App Servers...and where is SMP?
I don't think the BSDs will die, they will just be relegated to a class of specialized OSes as Linux continues to improve and consolidate it's position as the mainstream Server OS on the X86 platform.
threads?
You have 'linux companies' that have lost large parts of their market valuations, Linux distros merging, IPO's cancelled, etc.
In short, if one wishes to portray the 'worst' of the present state of the linux market, it would look like linux is the market that is dying. And, the total money gone in the linux market is far less than the BSD market. I'm betting you could get a link or 2 to the Brett Glass rants about how the GPL works to un-employ programmers. Given all the money that is gone in the Linux market never to be seen again....why not blame the GPL?
Go one better than the 'BSD is dying' person...provide actual links rather than their handwaving about BSD is dying.
Remember also that the BSD is dying troll had blown it in the past....declaring BSD dead due to Applixware not having a 5.X version for BSD.
It seems silly to propose the death of BSD as March 24 (the release date of Mac OS X) approaches. With one swell foop the installed base of BSD will grow by millions of machines.
The BSD license will always be attractive to companies who want a free code base unencumbered by a viral license - so as they say, I'm not dead yet.
MOVE 'ZIG'.
This effectively rules out any Linux 2.0-based systems. And considering the uptimes needed to get on the list (659 is the lowest when I'm writing this), Linux 2.2 has had very little time getting on the list. Should've been up since sometime around the end of april 1999 (somebody check this?). This allows for about 3-4 months from the release of linux 2.2.0, which means that the only admins to put this on their webservers would be freaks who just gotta have the latest bleeding-edge thingy running their server, and I'm very sure that these freaks have upgraded their systems since. Furthermore, this rules out any linux-kernel newer than 2.2.7 (which was released april 27, 1999) Now, if I remember correctly, 2.2.0 - 2.2.4 had *issues*. (and 2.2 didn't get wide acceptance before sometime after 2.2.10 or so... no wait... Red Hat released a version with some 2.2.5-version, I think... Of course, that proves nothing.)
Anyway, now some FreeBSD-lovin' wise guy wants to know why linux 2.1 isn't on the list of top uptimes. To this I can only offer three explanations: 1) web-admins are unwilling to run their servers on a kernel that is officially unstable. 2) officially unstable *means* unstable. 3) If you run an unstable kernel, you probably do it because the stable lacks support for some of your hardware. In this case, you're probably very inclined to upgrade to a newer stable kernel which supports your hw.
Don't wanna flame and, as I said, I make no claims regarding the relative quality of the two OS's, but those statistics prove nothing except that they don't prove anything. (Prove me wrong! ;-)
Now, if any linux 2.4.x-based system was on that list... That would be freaky!