Tim O'Reilly Confirms BSD Publications

InfoMonk writes: "I attended a library conference over the weekend. Tim O'Reilly spoke at a presentation on Open Source Software for libraries. After the conference I asked him about the long running interest in O'Reilly putting out BSD publications. He confirmed that two projects are currently in development, the expected BSD in a Nutshell and another book whose subject is not yet clear. This is very good news of course, to BSD hackers who are slightly tired of the press coverage that Linux has been given in the past year."

Re:How do the 3 BSDs relate to each other?

I think you're trying to start a flamewar, but that's irrelevant. The answer is: BSD4.4-lite. BSD Unix was born in the early Eighties, when the CSRG at Berkeley got hold of Unix. (Shortly after its birth at Bell Labs, many Universities were allowed to use it for research and educational purposes. Much of this development made its way back into the official code.) The CSRG did developed much of what Unix is today, from vi and the C Shell, to Unix TCP/IP.

Yes, TCP/IP. Back when ARPANet was being developed, DARPA needed an OS implementation of TCP/IP to work with. Thus, they funded Berkeley CSRG during the mid-late Eighties, and BSD is intricately tied to the roots of the Internet.

TCP/IP made its way back to AT&T, along with a lot else. Even with the commercial proprietary OS boom of the Eighties, there was much code sharing, and so the many Unices never got too far apart. Obviously the major reason for this is they all had to license it from AT&T. ;-)

System V, Revision 4 (SVR4) brought together the best of the Unix world, and while based largely on AT&T's UNIX, also includes a lot of BSDish stuff. Other companies' work was used as well, including pieces of Xenix. SVR4 is POSIX, what your beloved GNU/Linux is based upon.

SRV4 is the "modern" UNIX base, used in all commercial UNIX OSes. (Solaris, AIX, Irix, HP-UX, et cetera.) A note on Solaris: SunOS (as it was originally known) was formerly BSD based; not surprising, considering that Bill Joy is its father. But Sun switched to SVR4 in the late Eighties/early Nineties, and it has been known since as Solaris. Solaris 8 is really Solaris 2.8, which is SunOS 5.8. (Look at your Solaris bootlog...)

In the early Nineties, the original Berkeley CSRG was ready to call it quits. Before they did, they wanted to release their complete modern OS codebase to the community, as 4.4BSD. But there were still lingering pieces of UNIX code in BSD, which were removed after a series of messy lawsuits. The commercial code was rewritten, and the OS, now completely free, was released as 4.4BSD-Lite.

This codebase, whose history goes back to the very beginnings of UNIX itself, was adopted by several groups of developers who wanted to revive BSD Unix. The two most successful were FreeBSD, who concentrate on being the most advanced BSD Unix for the x86 arhictecture, and NetBSD, whose goal seems to be to run BSD on every processor ever designed. :) Not long ago, a disgruntled BSD developer forked the NetBSD code and created OpenBSD, an OS with tons of cool integrated cryto-nrrd stuff, which is famous as being perhaps the most secure network OS currently in use.

So, to the point, all *BSD OSes are based on the 4.4BSD-lite code which was the Berkeley CSRG's final work.

The BSD which you forgot to ask about is BSDi. BSDi is a commercial BSD company, whose board of directors includes a few original CSRG members (I believe). BSDi recently purchased FreeBSD's distributor and will be merging the codebases. The best news to come so far of this is that FreeBSD will finally have a native Java2 development kit.

---------///----------

Re:How do the 3 BSDs relate to each other?

All modern commercial Unices are SVR4. GNU/Linux, based on POSIX, is more SVR4-like, but is really a mongrel whose command syntax is usually a mesh of both, with stuff of its own added in. Regardless, the kernel structure is close enough to SVR4 to be considered such.

POSIX is the official standard which defines UNIX (as in the trademarked, licensed kind). A POSIX OS is essentially the same thing as SVR4. Close enough for this discussion, anyway.

Read my other response in this thread for more info.

---------///----------

Re:Linux v. BSD

From where I sit, there's no denying that the total number of BSD users is growing. Saying that "BSD continues to bleed market share" is another way of lying with statistics. For example, if the BSD market is growing at a 10% rate then the BSD market is pretty healthy, even if other market segments like Linux are growing faster.

Do you have any statistics at all that the number of BSD users is diminishing? If not then don't bring it up. This isn't television and the Nielson ratings. BSD won't get cancelled for being rated number two in the minds of newbies. Unix veterans do not abandon their favorite OS just because the Netwatch numbers are low.

Re:Linux v. BSD

I can see how a little resentment can creep into the most level-headed individual when things like political, religious or operating system choice are at play.

This "resentment" does not necessarily stem from ideological sources, or even disappointment that one's favorite OS is not number one. Rather, a lot of it arises from Linux bigotry. Hear me out!

It's extrememly frustrating when Linux is reported as being the only free OS. It's doubly frustrating when even those who know better say the same thing. For example, many applications written for X that use no OS system calls are advertised as "for Linux".

Walk into any given bookstore and the odds are high that you'll find open source related books under the heading of "Linux".

In terms of O'Reilly books, Tim said once upon a time that he would not publish a BSD book because it would be redundant. Yet there exists multiple redundant Linux books. How redundant is "Learning Debian GNU/Linux" coupled with "Learning Redhat Linux". For really eye-opening redundancy, check out the content of "Linux in a Nutshell" and "Unix in a Nutshell".

On Growth of the BSD Community

Well, I'd be willing to bet that most BSD people would welcome additional growth to the community. After all, BSD thereby gets more users, attracts more developers, and gains more/better software. With the increasing popularity of Linux -- or if I may, free UNIX -- I can only see growth as inevitable as *BSD gets swept up in the whole thing.

However, there's something to be said when you've got such a small, tight community as the BSD communities. Take the FreeBSD community, for instance. It's decidedly the largest group of *BSD afficionados out there, but it's still tiny compared to the Linux community. You can still get in touch with all of the main hackers. Greg Lehey, who wrote the book on FreeBSD (that is, literally the one printed book), still participates on several of the official mailing lists and answers when contacted by e-mail. I'm a complete nobody in the FreeBSD community, but I've talked (e-mail) with Greg many times.

Let us savor these things, us BSD people, because like it or not, the times they are a-changin'. I hope we can be as good big as we have been small

???

Walk into any medium-to-large bookstore in Japan. You'll find, n the Computer section, several shelves of Unix-related books. Among them, you will see many along the lines of "Build a Linux server", "Play with Linux", etc. etc. You will also see a nearly equivalent number of *BSD-related publications, everything from "Running MacBSD-68K" to "OpenBSD Security". There are beginners' books and weighty tomes on BSD kernel internals. And all this in a country with a population of half that of the US, and which is basically unable to export books to other countries because of the expense involved in translation (not to mention the "lag time" it adds).

My question is, "Why hasn't O'Reilly already published BSD books?" There's certainly enough potential interest to make it worth their while if they can put out books on Lego, f'Chrissakes.

Re:Linux v. BSD

These people are not stupid, they are using Linux because it fits the job they are deploying it for.

Of course they're not stupid. They use the OS best suited for the job. In some cases that will be Linux. In others it will be Solaris or AIX. Still others use BSD.

I really don't understand the point of this. What difference does it make to BSD if somebody is using Linux instead? Is this some sort of all or nothing warfare where there can only be one winner? BSD never set a goal for world domination. Such an idea is abhorrent to it. That eleetist attitude belongs to Linux and Windows.

What Linux lacked from the beginning is the eleetist attitude that some in BSD land are trying to get rid of.

You are right when you accuse BSD of having elitist subgroups. But you are equally wrong when you say that the Linux community does not have the same. Every operating system has its elitist and self righteous advocates. Some of the most elitist and arrogant OS folks I have met are Linux users. Just go to any Slashdot story about the release of any Linux distribution and you'll see it. Debian users have the only moral operating system, Slackware users have the one true distribution, Redhat users possess the real Linux standard, ad nauseum. They won't say it in so many words, but the elitist attitude is there and very visible.

Linux v. BSD

This is very good news of course, to BSD hackers who are slightly tired of the press coverage that Linux has been given in the past year.

I can see how a little resentment can creep into the most level-headed individual when things like political, religious or operating system choice are at play. Nevertheless, I see the mindshare gains acquired by Linux and Open Source over the past two years as a patently good thing for other- some, I'm sure, will argue, better- free Unices. A lot of organizations have, it seems, warmed to the idea that commercial software might not be only available option, and as they discover that viable alternatives exist to their, say, Windows-centric perspective, Linux will not be the only OS that will gain popularity simply because once shown the possibility of choice, one's point of view usually becomes quite a bit more flexible.

That said, the fact that a great number of closed-source ISVs only support Linux or even, commonly, one particular distribution, does irritate me immensely, as does the growing disregard for portability in Open Source software written by Linux users for Linux users.
--
Violence is necessary, it is as American as cherry pie.
H. Rap Brown

Re: Markets in a Nutshell (tm)

But they don't make this information public

Then this information is pretty much useless then, isn't it? Stop citing statistics that you won't release.

But even if you did release them they wouldn't draw the conclusions you do. Because you don't have the faintest clue what "marketshare" even means.

Is BSD losing? ... who is winning?

Marketshare has nothing whatsoever to do with winning or losing. This is not some Mayan game where the winners get to execute the losers. It ain't a game at all, especially not one of those zero-sum games beloved by the economically illiterate.

Let's look at a non-OS market like beverages Coca Cola and Pepsi have huge market shares while Royal Crown has very little. Does this mean that Royal Crown is a loser? Hardly. People who like Royal Crown cola will continue to buy it. And no statistics that could possibly cite about marketshare will ever convince them that Coke and Pepsi taste better.

But the OS market is not the beverage market. The marketplace for beverages is saturated. The only thing that can increase the total number of beverage consumsers is an increase in the total population. But the OS market is still growing. It is near to pointless to cite BSD's marketshare as an indicator of its death, when the total OS market is doubling every year or two.

Think about it. If the BSD market share lost 5% last year, they must be doing awesome! Because the total number of new OS consumers was much much more than 5%. BSD has more users today than they did last year. That's successful in my book.

In addition, lumping all operating systems together is erroneous. Just as Windows is targeting a different audience than Unix is, so Linux is targeting a different audience than BSD. Other than Debian and Slackware, the Linux distros seem to be targeting unix newbies and corporate users. The BSDs on the other hand are targeting unix veterans. They have never claimed to be the one OS for everyone.

The fact is this: as long as BSD continues to gain in total number of users, then BSD continues to gain in total number of users.

printed BSD manual

I would love to see the BSD manual published. O'Reilly breifly did that (probably almost ten years ago), but now it's all but impossible to find. I contacted them, and apprently they don't even have any copies in their warehouse. I'd love to have an updated copy.

But where will they get the manual from? CSRG is gone, and there are now three major BSD4.4lite-based OSes. My guess is they would go with FreeBSD's manual, since FreeBSD is the most successful, well known, and, some would argue, the most advanced of the three. But, truth be told, OpenBSD has an excellent manual, and I'll sometimes even consult it when I work with FreeBSD boxen.

Speaking of manuals, an excellent resource is the FreeBSD Project's manual webpage [freebsd.org], actually a CGI script which allows to you to consult manuals from CSRG BSD, FreeBSD, NetBSD, OpenBSD, several Linux distributions, SunOS (pre- and post-BSD-Exodus), as well as less common Unices such as Minix (har har), Plan 9, Ultrix, and Seventh Edition.

While we're ontopic, Sun Microsystems publishes a Solaris reference which is essentially the printed manual, but I would not recommend this book. All of Sun's Solaris books are absolutely horrid. (Which I've never understood; their Java books are wonderful.) Complete wastes of money. The only remotely useful things are the discussions of the mail system and network filesystems in the "Advanced System Administration" book.

Hands down, the best single Unix command reference book you can buy at the moment is O'Reilly's Unix in a Nutshell, but that is straight SVR4, with some Solaris-specific commands thrown in for good measure. You'll get what amounts to an abbreviated manual, along with stuff on shells, editors, and maybe a few system calls. If you would like to do system-level programming, Stevens' Advanced Programming in the UNIX Environment will counter-balance the Nutshell book nicely. Can never have too many reference books. :)

Forgive me, I love computer books, and tend to talk about them a lot

But that still does nothing for BSD users. Face it, guys, BSD is seen by much (dare I say most?) of the business IT world as "old" UNIX. Which it is, of course, but I don't find anything particularly wrong with that. But, it's a stumbling block which caused me to stop pushing it. (That, and the fact that my company is large enough to need the two dozen Sun and IBM boxen. I don't see many companies switching from Big Iron to clusters of x86 Free Unix boxen, but the high value and TOC of cheap x86 Unix are going to keep the current wave of IT startups going for another ten years.)

I honestly don't think I would buy BSD in a Nutshell. I'd rather have a printed manual, to complement Unix in a Nutshell, rather than replace it. Isn't that all I'd need? The vi and emacs syntax will be the same. So will awk, sed, and grep (for the most part). I'll be carrying two books around either way, so that stuff is just redundant.

Some say, "why have a printed manual at all, when everything is available on the system or online?" But you may be doing work at a location with both old and new SunOS boxes, and you're working on new SunOS, and you want to be make sure your shell scripts are portable between boxes. Having a printed BSD manual might come in handy then. Or you're on the subway going home from work, and you get in an argument with your geek buddies about proper BSD shutdown syntax. Or if you're like me, you just love computer books. So I'd buy the manual. But not the Nutshell book.

Okay, I'm ranting again.

"InfoMonk" mentions a mysterious BSD book in the works? I'm excited. I think a book on mid-level kernel work would do well. (Something to fill the void between The Complete FreeBSD and Design and Implementation of the 4.4BSD-lite OS.) Or a book on BSD-specific networking and TCP/IP. Hey, Roblimo, get us an interview with TOR so I can bug him about this. ;)

---------///----------

Re:Old news, no news.

I think this is largely due to the kind of people who tend to run BSD, compared to those who tend to run linux. While linux is admittedly installed in a wider variety of locations than BSD (IE, it is installed in the mission-critical spaces but is commonly used as "UNIX with training wheels) BSD tends to be run primarily by hardcore geeks and people who have work to get done.

What I'm saying here isn't that linux isn't suited to doing work, but I am saying that the sort of people who run BSD tend to be those who don't need or want a book. If you get on #linux you'll see a trillion clueless newbies hopping on and asking things like "How do I start a program at runtime" without ever poking their nose through /etc. By contrast, those who don't have a good handle on init scripts and the like are probably going to have so much trouble with *BSD that they are going to give up and go to linux.

I don't see this as a deficiency in BSD. It's a feature. BSD has been left fast and light. If that alienates some people, so be it. I didn't drag my way up from Xenix-286, through Slack and RedHat just to find that BSD had become bloatware. I know what I'm doing, and when I don't I know how to find out, by combing through manpages and poking about in /etc. I don't need documentation in twenty languages. Sometimes I need a book (Mastering Algorithms in Perl owns you) but at this point, I don't feel I need one to perform systems administration tasks. If need be, I'll poke through source and read comments and see what the variables look like throughout their lives. I can get it done.

I suspect many of you are the same.

How many people who commonly read the BSD news on /. are going to buy a nutshell book?

Re:How do the 3 BSDs relate to each other?

POSIX is an intersection between BSD and SVR4

Well, it's sort of the intersection between various flavors of UNIX, with some stuff added on; it antedates SVR4.

with BSD signals and terminal handling.

The signal mechanism is derived from that of 4.2BSD (and also happens to be the "main" signal mechanism in SVR4; other ones are supported as well). The terminal handling is more SV-flavored than BSD-flavored (where "BSD-flavored" really means "Research UNIX V7-flavored"; BSD added a bunch of additional stuff, most of which is also in SVR4, albeit with a more SV-ish flavor). Modern BSDs, and Linux, also have SVR4-flavored terminal handling (i.e, termios, with the UI extensions such as ECHOKE, ECHOCTL, word-erase and reprint and... characters).

Solaris is SVR4 as you say, but has a BSD filesystem.

SVR4 itself has a BSD-based file system (as well as, at least when it first came out, the V7-flavored System V file system - but, then again, 4.1BSD's file system was also V7-flavored).

I am still a bit puzzled about what you said about Linux: if it is SVR4,

Linux is Linux, just as HP-UX is HP-UX, IRIX is IRIX, AIX is AIX, Digital UNIX is Digital UNIX, etc.. Most of them implement POSIX, and a pile of other stuff, over and above POSIX, which is mainly the same on all UNIX-flavored OSes.

Linux isn't SVR4; most Linux distributions have a System V-ish init, although Slackware, I think, has a more BSDish init. Linux looks SVR4-like in a number of ways, but it's not SVR4.

and it is also Posix which signals does it have: BSD (as it is Posix) or SVR4? Or both (is this possible)?

At least on the RH 6.x (for some value of x) system here, there's a sigaction man page, which means it has POSIX signals; sigaction is derived from, and similar to, the BSD sigvec mechanism, but is not identical to sigvec.

There's also a sigvec man page on that Linux system, but it says

Under Linux sigvec is #define'd to sigaction, and provides at best a rough approximation of the BSD sigvec interface.

SVR4 and modern BSDs also have sigaction(); one should use sigaction wherever possible, and fall back on 4.2BSD-style sigvec() or "traditional UNIX"-style signal() (whose behavior differs between BSD and other systems; I'm not sure which behavior Linux gives, and I'm not sure how much I should care, as one should use sigaction() if one is writing portable code - yes, I know, there are probably signal() calls in Ethereal, I'll fix them) or 4.1BSD/SVR3-style sigset() only if one cannot use sigaction()

In other words, the correct answer to your question is "it has POSIX signals, just as SVR4 and modern BSDs do; those aren't exactly the same as BSD signals, but they're like BSD signals."

Re:an honest opinion

Well you might want to prove that. Please show me a Linux distribution (yes distribution, no gross kernel hacks) with a unified, clean source tree [netbsd.org] complete with all the platforms in the tree [netbsd.org].