A question of perspective

A question of perspective
by Greg Lehey

In mid-September I went to the Australian UNIX User Group (AUUG) annual conference in Melbourne. It was an interesting conference and I was not the only person left wondering if I was pointing in the right direction.

Last year's AUUG conference was in Sydney, and we had about 280 participants. This year we only had about 160. It was a great conference——I liked it better than the Sydney conference--but of course we had to ask why there were so few people present. Sure, the same old faces were there, but that was half the problem: the old faces were there, the young ones weren't. It would be easy to get the impression that young people don't use UNIX, but we know that's not the case. Obviously the problem is reaching new UNIX users.

So who uses UNIX nowadays? One way to answer that question would be to look at the user groups that do work. The most obvious ones I've seen here are the Linux user groups. In the course of this year we've seen a great increase in the number of BSD user groups and their activity. And AUUG is suffering.

The real problem appears to be the old UNIX habit of splitting into smaller groups. That's not bad, as I've pointed out before, but it makes it difficult for UNIX to appear to be a single system. If you buy Microsoft, you buy Microsoft. If you buy UNIX, you buy a particular vendor's implementation, or BSD or Linux. There's less obvious connection between the systems.

The same thing applies to the users. The Linux people get together at their user groups. The BSD people are starting to get together for their groups, either for a single BSD port, as in the case of SAFUG, or for all BSDs, as in the case of BUGS. None of them show any great inclination to join AUUG. In my state, South Australia, there are a total of 20 members of AUUG, and there's no local chapter.

So what's wrong with this? One of the considerations is that the AUUG is, indeed, a collection of old farts who are out of touch with reality. ``Linux is where it's happening''. We have a number of alternatives: We can accept the fact that UNIX as AUUG knows it is dying, or we can embrace the Linux people. We're still discussing what to do.

What does this have to do with BSD?

By now you're probably wondering why I'm even talking about this. This is a BSD publication, not a UNIX (System V) or a Linux publication. Well, one of the things that hit me during the course of the conference was how the three groups fit together. BSD is in the middle: on the one side is proprietary UNIX, with which we share a common history, on the other side is Linux, with which we share many open source concepts. It reminds me of the political system in Australia and many European countries: there's a large right-wing party representing the establishment, and there's a large left-wing party representing the workers, whom the establishment considers to be upstarts. Still, they have a lot in common. Both are relatively uniform in their approach. If necessary, their parties enforce compliance. One or the other of these parties wins a slim majority in elections, and they form the government, often without an absolute majority.

In addition, in the middle there's usually a small party or group of parties. Compared to the large parties, they're often unusually active and less categorical in their approach than the big parties. Often the big parties rely on them to be able to govern it all, so their influence is much larger than their size would suggest.

Sound familiar? It's easy to make too much of this similarity, but it's striking. BSD is the small group in the middle. We're smaller than System V or Linux, but we're more influential than our size would suggest.

Who cares?

So what difference does that make? I've found an interesting similarity, but what good is that?

To be honest, I don't know. It's something that we need to think about, and by writing this article I'm thinking out loud.

In politics, the people don't always vote for the best party or candidates. Biggest or most popular doesn't always mean best, and the majority doesn't always go for the best ideology. History is littered with examples of dangerous extremist parties being elected with large majorities. You just have to look at the state of the computer market today to see that this principle applies there as well.

For these reasons I don't think it's a good idea for the BSDs to try and go for market share: that would require changing from what we are. BSD has been a source of innovation in UNIX since 1980, when 3BSD, the first usable virtual memory version of UNIX, was finished. The Eighth Edition of Research UNIX (the original UNIX developed by Kernighan and Ritchie) was derived not from the Seventh Edition, but from 4.1cBSD. The main reason for this was the 4BSD had virtual memory support, while the Seventh Edition didn't. See the UNIX History Graphing Project for more details. You can also get this information in graphical form. This diagram also shows how much influence 4BSD has had on UNIX System V.

The Computer Sciences Research Group at Berkeley (CSRG) has closed down, but that hasn't stopped people from developing innovative modifications on BSD. If we were to merge with one of the other UNIX-like systems, this would be lost. Interestingly, this is a viewpoint that is shared in the Linux community. Jon ``Maddog'' Hall, the Executive director of Linux International, told me that he has, in the past, contributed money to a BSD project for exactly this reason.

So what difference does it make?

Even assuming that this interpretation makes any sense, you might wonder what difference it makes. Well, some of the more rabid BSD advocates are out to convince the world to drop any other operating system and move to (their version of) BSD. Other people say ``if they want it, they can come and get it'', and ignore the outside world.

I think that both these approaches are suboptimal:

The ``if they want it, they can ask for it'' attitude can discover that people don't really want things that much. One BSD has bitten the dust: 386BSD was the original Open Source version of BSD, but the last anybody heard of it was several years ago. Some kind of advocacy is necessary. Daemon News does its part to ensure that BSD is well known, but the individual projects need to do something for their visibility.

I don't think much of the ``let's rival Linux'' attitude either. In today's society, it's fashionable to want to be the biggest. In open source, that's represented by Linux. A number of BSD users are envious: ``We were there first, we're much better than they are, it's not fair that they're killing <my>BSD, we must fix that''.
I don't agree. The Linux people are doing a great job, as I've said in previous articles. In many ways, they're doing the donkey work for BSD: a lot of the public relations efforts for Linux also help BSD. Just recently a couple of high-profile newspaper articles were published, in the Wall Street Journal and the Boston Globe. It's not really surprising, of course, that after the press has written just about everything it can about Linux, it will look for new material, and that's what seems to be happening.

So what do we do?

So we don't try to beat Linux at its own game, but we don't sit back and do nothing. What do we do?

The most important thing I learnt at AUUG this year answers this question, at least in part: we need to recognize our position in the community, and to cooperate with other users of UNIX and similar operating systems. We're in a good position to do this: the BSDs are some of the most stable operating systems available at any price, and both the systems and their developers are highly regarded. We have a great contribution to make to the UNIX society, but the UNIX society also can help us.

Questions, not answers

But what is the contribution? That's the question I can't answer. It's one that we need to find, together, in the time to come. There are a number of things, though, that spring to mind:

We can contribute superior code. The BSDs have long had a reputation for high performance, not a traditional UNIX feature. The original Fast File System is the standard file system on UNIX, so much so that commercial UNIX changed its name to UNIX File System or ufs--not bad for software that isn't allowed to call itself UNIX. Nowadays, the FreeBSD virtual memory system, though still by no means perfect, performs better than most commercial systems. NetBSD runs on more platforms than any other system, and as a result it serves as an example for other systems when porting to new platforms. OpenBSD is performing a long and painful line-by-line analysis of the source code, fixing bugs which have been in UNIX for decades; most of these fixes apply to other BSD systems as well, and many apply to commercial UNIX and Linux.

We can contribute standards. Many of the current de facto standards grew up with BSD--just think of the Internet Protocols. In addition, the BSD teams have created development standards that rival the best commercial practices. The XFree86 and Samba teams have adopted much of the BSD development model, including the CVSUp program for automatic replication of the CVS repository.

We can explore alternatives. Consider the issue of software RAID and volume managers, which interest me particularly. Linux has a Volume manager and a separate RAID implementation. FreeBSD has Vinum, a volume manager which includes RAID-5. NetBSD and OpenBSD use RAIDframe. Which is best? Time will tell. If we had only had one of them, there would have been no way to know whether it was the optimum. In addition, the competition makes people write better code.

But couldn't Linux do this without BSD? Of course they could, but they don't. There's only a certain amount of effort to which any one team will go, and there are enough challenges in any project without reinventing the wheel.

What can the UNIX society do for us?

A lot of this looks very much as if BSD users ``go it alone''. Yes, I think that's a correct impression. That does duplicate a lot of effort. In the previous section I talked about a situation where it makes sense. In many cases, it may not. People are talking about <my>BSD certification, both for products and for staff. It's difficult to get vendors to go through the compliance business for half a dozen different operating systems; even finding space on boxes to place all the logos would be a problem. It would seem to make more sense for the BSDs to join forces not only with each other, but with Linux: most of the certification issues are the same.

The Linux people wouldn't buy into that, you say? Maybe not, but I did some preliminary investigation at AUUG, and the signs look positive. At the moment the main problem seems to be the reluctance of the BSD camps to embrace anything that wasn't invented here.

Where do we go from here?

As promised, there aren't many answers in this article. The Open Source community is changing rapidly. Most of the impetus is coming from the Linux community, and there's a certain resentment in the BSD community as a result. We shouldn't let that cloud our view, though: we have interesting times ahead, and it's up to us to take the best advantage of them.