herbie@polaris.UUCP (Herb Chong) (02/25/86)
I just found out that for the past month or so, all news articles posted from polaris have been trashed so none of these (maybe) made it out into the world. bear with me if these are repitions. Herb... ================== Subject: Re: unix brings back obscurity? References: <2457@ukma.UUCP> <12173670725.8.MRC@PANDA> In article <12173670725.8.MRC@PANDA> MRC%PANDA@SUMEX-AIM.ARPA (Mark Crispin) writes: > Are you aware that Dennis Ritchie once said that if he had >known about Tenex, he never would have invented Unix? The goal in >designing Unix were to create an quick and dirty operating system >on a discarded PDP-7 so a group in Bell Labs (which had been denied >their purchase request for a newer and large CPU) could get some work >done. Steve Bourne spends about a page in the introduction (or preface, i forget which) describing the history of unix. it was a hack, pure and simple. they had limited hardware, so they tried to do as little as possible to get a system that supported what they wanted. with limited objectives, they succeeded. it so happened that the design was sufficiently limited that it could be moved to other machines with little difficulty, especially after the system was rewritten in C. everywhere possible, a limited, simple, and usually slow, algorithm was chosen over a faster and more complex one. unix represents in a sense, a system that assumes the least possible about a hardware configuration, and so is easy to port, but that also means that it can be highly inefficient. a unix filesystem is fragile, and there is a tremendous amount of programmer intervention required to do some things. it is poorly documented and obscure in many ways. many people have used the excuse that the systems come with source to provide less than useful documentation. in short, a hackers dream. on the other hand, V7 and other systems larger than it come with a larger than typical set of program development tools to start with. many are riddled with bugs and others require prescience to understand what is happening, but at least they are available. the success of unix has little to do with how good it is but more how easy it was to port to another machine. there is also the pretense of a "standard" unix too, but in practice, the standards are less than rigid. not that i can't stand unix, but there are many short-comings to it that i don't have the time or energy to fix (yet). it has a lot of good ideas, some implemented well, and others implemented poorly. it needs work to make it run on bigger machines supporting a proportinately larger number of users. the filesystem needs to be less fragile. the documentation needs improvement. still i use it, and for some things, it is the appropriate choice. OOOPS! forgot to mention which book. The Unix System by S. R. Bourne. Subject: Re: unix brings back obscurity? References: <12173670725.8.MRC@PANDA> <8601091624.AA03662@pc.Purdue.EDU> i started to mail this only to find that there was no return address to mail to except the info-atari mailing list on ARPA. apologies for spreading this all over the net. >>Nor >>is it "efficient", since Unix systems typically spend about 50% >>of the CPU in the operating system. > Who cares? CPU time is cheap, programmer time is not. a common fallacy. CPU time is cheap if you want lots infrequently. if used often enough, any code can stand some improvement. in the case of most unix implementations, it is the most frequently used code that needs the improvement. besides, upgrading from a VAX 11/780 to a 785 is not cheap. if reducing system time by 10% saves me a couple of thousand dollars for a few months, it would have been worth it. so alright, for an ST, the difference is trivial, but unix is not IT, folks. it may be better than a lot of it's competitors, but that isn't saying a whole lot. >>One questions the flexibility >>of an operating system which lacks file locking, record structure, >>and shared writeable pages -- granted these are now being shoehorned >>into Unix, but they should have been in there from the beginning. > Why should they have been in there since the beginning? Some of the > most powerful database systems in existence run under UNIX. They seem > to get by just fine. What type of record system would you like? One > like VMS, one like RSX, one like MACE? And what happens when you > want to move software between the systems? Some people feel that these > type of things do not belong in an operating system; I am one of them. define powerful database system. raw throughput is number one thing that a unix system is not very good at. having to build all your own tools is another. there may be good database systems out there for unix, with lots of capability, but if it drives a system to its knees running this wonderfully flexible command language, search function, or report generation, then you may want to make a whole lot of tradeoffs to get the speed you need. an OS that gets more out of the CPU may be the prefered choice. as for why the tools should be there, why should everyone have to reinvent the wheel for their applications? it's like someone asking me for a wrench and i hand them a bar of steel. they can make any kind of wrench they want out of it, but do they have the time? unix is so flexible because it does nothing (nearly) for you in terms of filesystem structure. conversely, you have to make your own tools all the time. >> >> Are you aware that Dennis Ritchie once said that if he had >>known about Tenex, he never would have invented Unix? > So? it means that he himself would rather forget that he invented unix, despite its success. i have heard that he has once said that he feels like someone who started a religion that he now sees all the flaws in, but no-one else seems to want to listen. he feels caught up in something he no longer believes in. of course, if i'm misquoting, i'll retrack my statement, Dennis. >>The goal in >>designing Unix were to create an quick and dirty operating system >>on a discarded PDP-7 so a group in Bell Labs (which had been denied >>their purchase request for a newer and large CPU) could get some work >>done. > And the Greeks invention of geometry was so that they could settle > land disputes. That doesn't make the invention of geometry any less > significant: "Necessity is the mother of invention." That's not > original, but it is relevant. and some of the worst code in the world was written that way too. do you think that unix kernel code is easy to read? maybe in the V6 or V7 days, but not any more. and the greeks did not invent geometry, they formalized it. unix was hacked together to do something until they had something else to do it right. that right thing never came along and so more and more got added to unix. i'm disgusted at the people who think that unix is the best thing since sliced bread. yes, it does do a lot of things. yes, you can shoehorn almost anything onto it. but, yes, it's a hack and for hackers. since i'm a hacker, i like using unix for program development, but that doesn't blind me to all it's faults. ever try to develop and test a new kernel? you need your own private machine for that. VM/370 lets anyone do kernel hacking while a whole pile of other users are using the same machine for something totally different. does that mean VM is great? nope. but it does something well that nothing else out there does. unix is uniformly mediocre. it uses the lowest common denominator between a lot of different types of machines. in doing so, it doesn't try to do too much and it succeeds well at not doing too much. people have just gotten used to it, that's all. Herb Chong... I'm still user-friendly -- I don't byte, I nybble.... VNET,BITNET,NETNORTH,EARN: HERBIE AT YKTVMH UUCP: {allegra|cbosgd|cmcl2|decvax|ihnp4|seismo}!philabs!polaris!herbie CSNET: herbie.yktvmh@ibm-sj.csnet ARPA: herbie.yktvmh.ibm-sj.csnet@csnet-relay.arpa, herbie%yktvmh.bitnet@wiscvm ======================================================================== DISCLAIMER: what you just read was produced by pouring lukewarm tea for 42 seconds onto 9 people chained to 6 Ouiji boards.
henry@utzoo.UUCP (Henry Spencer) (03/02/86)
> ...a unix filesystem is fragile... This is superstition, not fact. It hasn't been true for quite a while. > the success of unix has little to do with how good it is but more > how easy it was to port to another machine... Unix was quite successful on the pdp11 before it was ever ported to anything else. Which was why it got ported to other machines in the first place. > ...it needs work to make it run on bigger machines supporting a > proportinately larger number of users. It's been done. Ask Amdahl. > the filesystem needs to be less fragile. It's been done, long ago. > the documentation needs improvement. Still a problem area. Although the documentation for most other systems also needs improvement, for different reasons: it drowns you in verbiage and detail. One thing Unix does *not* need is documentation that is "improved" to that point. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
nather@utastro.UUCP (Ed Nather) (03/03/86)
In article <6452@utzoo.UUCP>, henry@utzoo.UUCP (Henry Spencer) writes: > > the documentation needs improvement. > > Still a problem area. Although the documentation for most other systems > also needs improvement, for different reasons: it drowns you in verbiage > and detail. One thing Unix does *not* need is documentation that is > "improved" to that point. > Henry Spencer @ U of Toronto Zoology That "verbiage" that is so opaque often consists of passive constructions which totally obscure the point to be made. As an example, from the Microsoft C (v3.0) manual: "If more than one flag is present, they are separated by the "|" symbol ..." By whom? Me, or the compiler? After experimenting, I found it *should* say "If you include more than one flag, you must separate them with ..." This type of construction is depressingly prevalent in documentation of all kinds on all systems. It should be taken out and shot. -- Ed Nather Astronomy Dept, U of Texas @ Austin {allegra,ihnp4}!{noao,ut-sally}!utastro!nather nather@astro.UTEXAS.EDU
paul@whuts.UUCP (HO) (03/04/86)
> ... unix is uniformly mediocre. it uses the lowest common > denominator between a lot of different types of machines. in doing > so, it doesn't try to do too much and it succeeds well at not > doing too much. people have just gotten used to it, that's all. > > Herb Chong... > So, there you are Herb! Nice to find you at IBM, instead of at Burger King. Ok, Herb. In your article, substitute unix with IBM. Paul Ho