From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Sun, 14 May 2000 17:29:14 -0400 From: geoff@x.bell-labs.com geoff@x.bell-labs.com Subject: [9fans] open sores (was My view of Plan 9 and it's future) Topicbox-Message-UUID: ad65a0bc-eac8-11e9-9e20-41e7f4b1d025 Message-ID: <20000514212914.XOiUKiGAfmNxK5loSKGHeumtcRAALf4sz76DnFJKbRs@z> I am not a member of the Computing Sciences Research Center, where Plan 9 was and is developed. These opinions are not official corporate statements. I'll second Russ's and viro's comments. Mr. Choate, you're not paying attention, and you're apparently letting Open Source ideology drive you to say some pretty absurd things, often contradicted by history. (Your middle initial wouldn't be "N.", would it? :-) Open Source ``technology''', that is, letting everybody stick a finger in the pie (or a ladle in the broth), is no guarantee of quality software and I'd argue that it's quite the opposite, despite Eric Raymond's `The Cathedral and The Bazaar' or whatever it's called. Go dig up a copy of TMNN (B News 3.0), the GNU/Linux cat command, or the Linux Soundblaster driver if you want to see some pretty wretched code produced by Open Source methods. The only GNU software I've ever had any respect for is groff, which was written by one person (James Clark, in England), not the Cambridge GNU hive mind. At least the folks at Berkeley who produced the BSD distributions had the benefit of access to the original well-written Unix source. The people writing GNU/Linux knock-offs have generally not had access to Unix source and so are, at best, learning from reading BSD-Lite source or if they are really unlucky, other GNU/Linux source. Open Source seems implicitly to assume that all programmers are of roughly the same skill. This is obviously false, as observed by Gerald Weinberg in The Psychology of Computer Programming decades ago, among others. The Plan 9 developers are exceptionally talented, experienced and modest as designers and implementors. As a result, they have received various medals, awards and other honours. I hope that they will pardon the blatant flattery. To take a few examples, Rob Pike seems to have rare, possibly unique, insight into window systems and graphics, not to mention an unusual ability to distill software into quite compact form. With the possible exception of MGR, I've never seen a window system written by anyone else that approaches any one of Rob's for usability, compactness or elegance. Wiser people than I have said of Ken Thompson (quoting from memory): `Michelangelo would have had to have been taught how *not* to sculpt. So it is with the great programmers.' Ken's insight into file systems is penetrating. The number and severity of file system problems with which the maintainers of other systems are still grappling, and that Ken side-stepped in the Plan 9 file server, is impressive. And of course he created little things like Unix, B, the Plan 9 C compilers, and at least some of 9P. The skills of the Plan 9 developers extend beyond design; you don't get the full benefit of their abilities by reading the Plan 9 documentation. Writing and debugging software, especially kernels, for multiprocessors is harder than it looks. Now you *could* try to write a Plan 9 clone from scratch (let's call it Battlefield Earth) but even if you somehow attracted fairly good programmers, and they were able to spend all their free time (if any) working on it for free, it would be hard to justify the effort on economic grounds, so presumably the justification would be `freedom', even compulsion, to give the software away. When Unix source licences were at least $20,000 for non-educational uses, and Minix and BSD-Lite and Plan 9 and even Linux didn't exist, writing freely-distributable commands and kernels (even if only clones of Unix ones) had some utility. At $350 for the real thing, I don't find this so compelling. The VSTa people have already tried to write a clone, but I gather progress has been slow, which isn't surprising: spare time is scarce, PC hardware is vile, there's a lot of code to write from scratch, and if you take shortcuts like using gcc instead of writing 2c, etc., you reduce the quality, utility and certainly compactness of the result. The 1995 Plan 9 release was a bargain by any measure. The cheapest Unix source licence that I know of was a V6 educational licence for $500 in 1975 dollars; V5 may have been cheaper. And that was 25 years ago, before there was any significant networking or graphics code in the standard Unix distributions. The open sourcers who whinge about the expense have nevertheless somehow acquired computers that have until recently cost one or two thousand dollars, possibly more. They may even own multi-thousand-dollars cars (shock, horror). As Russ said, Plan 9 is a research operating system and as pleasant and productive as it is, Lucent (not AT&T) is under no obligation to make it available to anyone. I think that the terms under which Lucent has offered Plan 9 to the public in the past have been very generous. Twenty-five years of operating system expertise were made available in the form of a very usable operating system, with source, for $350. Admittedly, Linux costs less, but in this case at least, you get what you pay for. And, yes, it's galling that Windows outsells Linux, which outsells Plan 9. It could be a coincidence, but the evidence seems to support the observation that quality and popularity are inversely related. Pardon the vitriol; I guess I've been more annoyed with the know-nothing arrogance and sense of entitlement of some GNUers and Open Sourcers (Sorcerers? apprentices?) than I had realised.