Old School Linux Remembered, Parts 0.02 & 0.03 163
eldavojohn writes "Following our last history lesson of Linux 0.01, the Kernel Trap is talking about the following announcements that would lead to one of the greatest operating systems today. A great Linus quote on release 0.02 (just 19 days after 0.01): 'I can (well, almost) hear you asking yourselves "why?". Hurd will be out in a year (or two, or next month, who knows), and I've already got minix. This is a program for hackers by a hacker. I've enjoyed [sic] doing it, and somebody might enjoy looking at it and even modifying it for their own needs. It is still small enough to understand, use and modify, and I'm looking forward to any comments you might have.'"
Re:15 years later... (Score:4, Informative)
Great comment. Yeah, the Hurd and DNF basically had the same problem: they were great big-picture ideas, but actually implementing them turned into giant software development nightmares. In the Hurd's case, the main problem was that the GNU folks decided to base it on Mach. Then when Mach turned out to have major architectural problems, they moved on to L4. Progress on porting to L4 went slowly. But then, even that tack fell through as well, and so they became basically stuck as they are now. They're looking into several different microkernel bases - I think the one they're looking at now is called ?Coyote? Anyway, both DNF and the Hurd are sad tales of where concept doth not meet reality, and never the twain shall meet.
We're drifting _way_ off-topic. Let's remember Linux now, the major kernel that started out as a hack. It still is, somewhat, (in terms of the way it's sort of in between Windows and the concept of the Hurd) but it's now one of the biggest server OSes on the planet.
I shut up now.
Debian (Score:3, Informative)
The Linux alternate history game... (Score:5, Informative)
There's a lot to be said for the microkernel architecture, and if Moore's Law ever does start to level off, then I think we're going to see a move away from monolithic designs for good. It's just not practical to keep stuffing more features into a monolithic kernel if you're not constantly getting more and more memory to run it on, and only a very small body of users can be expected to ever compile their own. (True, you can always recompile a specialized version of a monolithic kernel, ripping out all the stuff you don't need, but this is a PITA and it only becomes harder as the thing gets bigger.)
Along with probably most other Linux users, I've always wondered how things would be if Tanenbaum had released MINIX under a free license earlier in the game (Torvalds has said at several points that had MINIX been more free, he probably would have simply modified it, keeping its architecture, but since Tanenbaum had no interest in "turn[ing] MINIX in BSD UNIX" [1]...the rest, of course, is history.)
Or perhaps more interestingly, what would have happened if a free version of BSD had been produced for low-end hardware just a little earlier than it actually was. (In reality, 386BSD came out in a working form in July 1992 [so sayeth Wikipedia], nine months after the first Linux release, and 4.4BSDLite didn't come out until '94 [2].) It seems to me that had "real UNIX" been available for low-end systems in the early 90s, much of the impetus to create a from-scratch clone would have disappeared. (Although, maybe not; perhaps the philosophical differences that drive Linux and the BSDs in different directions would have eventually caused a from-scratch rewrite.)
Ultimately I don't think either alternative would really have brought us out at much of a different place than we are right now, at least from an end-user's perspective; the majority of users don't really care about kernels as such anyway. But it's always fun to play 'what-if,' as long as one keeps in mind that although it's easy to fixate on how things could be better, it could always be far, far worse.
[1] Great archive of Torvalds / Tanenbaum Usenet discussions here [fluidsignal.com]. There's so much ego going on there, from both sides, ASCII text can barely contain it...
[2] I'm partial to fellow Slashdotter connorbd's BSD History [geocities.com], which is a good primer.
Re:The Linux alternate history game... (Score:5, Informative)
It's important in that performance was what killed the Microkernel. After Mach came up with such pitiful performance numbers, all the OS developers* who had been looking to embrace microkernel architecture ran the other way. At the time, the performance vs. security/stablity tradeoffs weren't worth it. Now with ultra-modern machines running on an insecure internet, things are *starting* to turn around.
* There were a few exceptions, of course. NeXT adopted Mach regardless of the performance issues. They spent a lot of time upgrading it into the hybrid XNU kernal. QNX was a microkernel because the guys who wrote it were very clever and didn't know any better. NT shows the basic design concepts behind a microkernel, but all the servers ended up getting shoved into kernel space for performance. So NT just barely scrapes by as a hybrid.
Re:The Linux alternate history game... (Score:2, Informative)
In the case of the Linux kernel, what I said is what it is good for.
Re:"Greatest operating systems today"? (Score:1, Informative)
95 = 4.0
98 = 4.1
ME = 4.2 or 4.5 (can't remember)
2000 = 5.0
XP = 5.1
2003 Server = 5.2
Vista = 6.0
You = Moron.