The Walnut Creek, BSDI Merger

Perhaps even more important than the Microsoft ruling this month is the merger between BSDI (makers of the commercial BSD/OS) and Walnut Creek CD-ROM, the financial backbone of the FreeBSD project. Actually, BSDI acquired Walnut Creek, as I understand it.

Why do I think this is more important than the Microsoft ruling? Because the ruling comes as no surprise. It would have been a surprise if it went in Microsoft's favor. (Although, I doubt the DOJ will over play the strength of its hand. It will try to not bully Microsoft into bite-size business groups.) Besides, Microsoft will probably restructure soon anyway, and they're smart enough to find another way to capitalize on the open source movement. And on the ever-changing nature of technology.

So, the BSDI merger is more important than the Microsoft decision a) because Microsoft will fight it for years, and nothing big will happen directly because of it right away, b) Microsoft has already been restructuring itself, what with Bill Gates once against assuming the role of chief technologist (just like in the old days), and c) the BSD camp has remained more purely open source than Linux, for example, because it doesn't restrict implementation of its source code even in closed source or proprietary projects, as does the GPL.

(NOTE: This does not imply in any way that either is ``better'' than the other. I'm merely pointing out the biggest difference between the BSDL and the GPL. I'm using ``pure'' not to imply ``worthiness'' in any way, but only ``less legally restrictive.'' The purest license I've run across so far is the Poul-Henning Kamp's ``Beerware License.'' See http://www.freebsd.org/~phk.)

``Purely Open'' BSDs. Hopefully, I've averted the flamage that comes to someone who claims that ``My [insert project name] BSD is better than Linux,'' because that's not what I'm not saying. This merger interests me because it's different from the Linux IPOs and spin offs and upstarts we've heard about in every corner of the media. It's a part of the open source community we don't hear that much about.

In contrast to Linux distributions, Berkeley UNIX implementations have not sprouted up like poppies. The four stalwart names are FreeBSD, OpenBSD, NetBSD, and the commercial BSDI. (Other projects include TrustedBSD, PicoBSD, Darwin, e/BSD, RTMX/OS, and Eclipse, some commercial, some free.) Some of the BSDI founders were the original contributors to the BSD Unix developed at CSRG on the Cal Berkeley campus. Only now it appears that BSDI and FreeBSD will merge into one free distribution. This is the opposite direction that Linux distributions have taken in the face of financial success. They spin off new distributions, not consolidate them. So what's going on here?

For one thing, perhaps there's not as much rivalry among BSD camps as Linux advocates and talking heads in the media sometimes claim. There have been some conflagurations in the past, including lawsuits, flame wars on Usenet lists and personal feuds between strong-minded individuals. News of these spats travels quickly in a small community. And it's easy to overestimate their importance, much as personal disputes between tribal elders can be upsetting to the lay members.

Despite the personal differences between some of the individuals, the BSD development model has remained a consensual democracy, in most cases, whereas Linux has enjoyed the benefits of having a skilled and benevolent monarch by the name of Linus Torvalds. This monarchy reaches primarily as far as only part of the kernel itself, however. Beyond that, there are projects and distributions galore, often employing more of a consensus-based decision making model like BSD's. However, as Linux has become enormously popular and well funded, the area that Linus is willing to oversee has grown smaller. Most Linux projects share the same sort of decision making process as FreeBSD does. And someone like Jordan Hubbard (one of the founders of FreeBSD) has much more in common with Linus Torvalds than not.

The source trees for FreeBSD and Linux are different, however. For example, FreeBSD has a core team who have CVS commit access to the source tree for the whole distribution, not just the kernel. Other BSD teams have somewhat similar teams. They don't have a Linus whom they allow to decide the ultimate fate after hearing all the arguments. They decide amongst themselves for the good of the clan, may the best hacker win, and then still love each other enough to want to work together tomorrow. It's not an easy challenge, but no single person has won such complete trust from others as Linus has among his peers. But again, I suspect that this difference between BSD and Linux decision making is weaker nowadays than when Linux was newer and smaller. I'm not an insider, so I can only listen in with my glass against the wall and guess from what I hear.

But what's interesting here with Walnut Creek and BSDI is that this will be the first merging between a ``free'' source tree (FreeBSD) and a ``commercial'' source tree (BSD/OS). And that it's happening amongst the ``red daemons'' and not the ``happy penguins.'' It will be interesting to see what happens and how.

Can Sources Merge? It appears BSDI and Walnut Creek have not decided how much to merge, if at all, their respective source trees. Certainly they will share certain parts of them, as they already borrow now from each other. But they don't appear to be hell-bent on making just one BSD composed from FreeBSD and BSD/OS.

There will be something called an Application Binary Interface originating from BSDI that will make it easier to port and create commercial (and free) applications that will (hopefully) support all of the BSDs. This is supposed to be a very good thing. One BSD application will more likely fit all BSDs if this succeeds. Each of the BSDs may profit from the efforts of another. However, other BSDs than FreeBSD seem to be taking a ``wait and see'' attitude to this innovation for now.

BSD/OS has several components that were developed under NDA (non-disclosure agreement) and that cannot be released as open-source. I've read that BSDI claims to share the BSD/OS source tree with core FreeBSD developers (http://www.daemonnews.org/200004/interview.html). To what extent? I don't know.

The acid test will be whether BSD/OS's SMP (symmetric multiprocessor) support gets rolled into FreeBSD. It's better than the FreeBSD's and would be like sharing the crown jewels with the in-laws.

Obviously, when parts of BSD/OS become part of the FreeBSD source tree, that part of BSD/OS becomes open source. It's unclear how much more of the BSD/OS will become open, or when.

BSD/OS has benefitted from the other BSD projects, however. The BSD/OS sparc port was developed largely from NetBSD, for example. I haven't heard of NetBSD or OpenBSD receiving any sources from directly from BSDI, though. If they get it, it must come through the FreeBSD source tree.

The New Company. The new company, according to Greg Lehey, a FreeBSD developer and tribal elder, will be called BSDi. That's a lowercase ``i'' following ``BSD.''

Now, what will happen to Slackware Linux, also marketed by Walnut Creek CDROM? According to Greg's article in Daemonnews (http://www.daemonnews.org/200004/dadvocate.html) it will not be sold any longer by the company. They will spin off a new company to handle Linux and other products (such as Hobbes OS/2 collections, I gather).

The BSDI interview posted in Daemonnews quoted above indicated that BSDi will market BSD as a whole. Not just FreeBSD or BSD/OS. I'm not sure how they can do this. I'm not sure how NetBSD and OpenBSD will benefit directly from BSDi's marketing. It sounds like managerial BS rather than BSD. But, if FreeBSD and BSD/OS are marketed well, NetBSD and OpenBSD will benefit indirectly.

What will really help NetBSD and OpenBSD are if BSDi funds some development on core areas as well as shares sources on already excellent components, such as SMP.

In short, I see the Walnut Creek/BSDI merger as a wedding between two well bred families. I eagerly await the birth of their progeny.



Last modified: Thursday, January 03 2000 (???)

dsj@dsj.net