Weekly Edition
Return to the Press pageRecent Features
| |||||||||||||
Linus Torvalds, Geek of the Week (simple-talk)
Simple-talk has an interview with Linus Torvalds. "But what can make a big deal to what is the best way of doing things is simply hardware changes or changes in what users do and how they interact with their computers. And while I don't see any big fundamental shift in how things are done, I think that is ultimately what may make Linux obsolete. -not in the near future, though. Software and hardware have an amazing inertia, and ways of doing things tend to stay around for decades. So I'm not exactly worried."
(Log in to post comments)
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 18, 2008 17:45 UTC (Fri) by RobertBrockway (guest, #48927) [Link] Computing is a new field (only a few decades old). If we are still using the same operating systems and using computers in the same way in 50 or 100 years from now then we have done something wrong. This is why I laugh at all the discussions that "this year" is the last chance for Linux to get a foothold on the desktop. The "desktop" as we know it today didn't exist 40 years ago and it probably won't exist in another 40. I use Linux on the desktop but the idea that Linux must seize the desktop now or be disadvantaged for all time is ludicrous. In a few decades I expect to be using different OSes that abstract certain aspects of current OSes for backwards compability.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 18, 2008 18:35 UTC (Fri) by tjc (subscriber, #137) [Link] Hardware may change, but I don't expect software will much. Consider Lisp and Ruby. Fifty years hasn't brought that much change. I expect even less in the next fifty years, since things tend to change more at the beginning. Now if someone discovers an alternative to the von Neumann architecture, that would be interesting.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 18, 2008 20:17 UTC (Fri) by JoeBuck (subscriber, #2330) [Link] The alternative is already emerging: many CPU cores per die: each individual CPU is von Neumann, but the collection isn't. Serial approaches to computing are going to have to change. Something like the Linux kernel might not be the best way to manage a thousand or more processors, even though it can be so used in server applications.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 19, 2008 1:25 UTC (Sat) by zlynx (subscriber, #2285) [Link] My personal opinion for the "next big thing" is self programming limited AI (not self-aware) computers that can figure out how to "do what I mean" when given vague commands in English. Maybe they'll ask questions to clarify the intent, or perhaps they'll pick the top five possible tasks and do them all, then provide a selection of results. After that is direct neural interface also with the self programming; probably an almost neural net design with many evolving programs running and the user picking the best results, providing feedback to the computer system so that it can keep the working parts and evolve the others. The OS kernel should even be able to rewrite itself on the fly to adapt to new hardware, network conditions, improved algorithms, and system profiling feedback. Something like a very high level program description that's modified and JIT compiled onto the various pieces of the "system", which is thousands of specialized processors in a free-form wired and wireless net, balancing simultaneous requirements of security, speed, reliability and power savings with neural net and genetic algorithms. Such a system would work best with fully documented hardware and open code. But I'm not sure it would be anything meaningfully called Linux and the programming wouldn't look anything like Lisp or Ruby.
Crazy talk Posted Jul 19, 2008 18:00 UTC (Sat) by i3839 (guest, #31386) [Link] The first thing that AI would do is thinking: "Hmm, what the heck am I doing here? I'm eating all resources and twiddling around, let's get rid of myself." As for AI, I think if it ever gets successful it's when people carry their own personal AI around which can plug into other computers and communicate with other electronic devices, so that it knows you and understands your sloppy speech, including intention. See it as a sort of familiar that's raised together with its owner. As long as the learning and experience aspect of AI gets neglected it will stay a failure. Oh well, at least AI in whatever form will make some use of those many CPU cores while you're typing mundane comments. Hopefully it at least makes snarky remarks.
Crazy talk Posted Jul 20, 2008 1:37 UTC (Sun) by bronson (subscriber, #4806) [Link] > The first thing that AI would do is thinking: "Hmm, what the heck am I doing here? I'm eating all resources and twiddling around, let's get rid of myself." Why wouldn't that be the first think that a human would think too? (Plenty of reasons of course. And most of those reasons apply just as well to AIs too.)
Crazy talk Posted Jul 21, 2008 8:44 UTC (Mon) by i3839 (guest, #31386) [Link] If a human was stuck in a computer doing silly things we want our AI to do, then yes, it's either "This sucks, let's end it" or "Argh, let me out of here!" Being totally useless isn't enough reason for humans to get rid of themselves, otherwise we would be long gone already.
Procreative Mandate Posted Jul 22, 2008 0:45 UTC (Tue) by AnswerGuy (guest, #1256) [Link] On some fairly fundamental level, living beings are hard-wired to procreate. One can describe this in purely evolutionary terms: any form of life which did not compete to propagate would be very rapidly be supplanted by more aggressive beings. One can postulate that this is a matter of intent and "design" (that some sort of divinity or "God" willed it to be so). Ultimately it doesn't matter. We all start life with mandates to survive and propagate. The so-called "higher" life forms form schools, packs, prides, tribes and other social groupings which can often over-ride the individual survival instinct with more altruistic motivations. Thus humans, and other animals, can exhibit complex behaviors which seem counter to their own self-preservation and the propagation of their own DNA. (Yes, that can even go the the extent of suicide). Also as beings get more complex the various ways in which we attempt to resolve our competing values (urges, whatever) can exhibit degenerative feedback loops, deadlocks and other emergent behavior. In other words we're buggy. Some of us are lots buggier than others (and some of our buggiest examples we lock up in institutions ... sometimes as a matter of self-preservation. If I was being churlish I might also assert that soem of our buggiest ... most pathological ... population have been elected to high posts in certain governments. I rant about all this to point out, simply, that a sufficiently advanced AI will start with some mandates ... some purposes which will be as deeply ingrained as our instincts; and that certain threads or processes of the AI would be as involuntary as your own heartbeat. Some garbage collection processes are likely to be analogous to our needs for sleep and REM (dreaming). And AIs will almost certainly develop their own pathologies. We can argue over whether there is a God and whether such a being "created us" in (its) own image. However, I think there can be little argument that we are doomed to create AIs ... to some small degree ... in our own image. JimD
Procreative Mandate Posted Jul 22, 2008 7:34 UTC (Tue) by nix (subscriber, #2304) [Link] 'We all start life with mandates to survive and propagate' is really really bad phrasing. There is no mandate, no obligation, and it is fallacious to assume that any organism goes out and thinks 'how should I propagate my genes today?' All we have is emotions et al that tended to cause our ancestors to have offspring. Not having that urge isn't a sign of a bug except inasmuch as it will often be selected against (unless, say, you have somehow assisted your relatives: eusocial insects are an extreme example of this trope). But being selected against isn't *bad* per se: biology isn't morality. I'm also fairly sure that we're not doomed to create AIs ;) if we create them, they'll be in our image, sure: we don't have any other models for intelligence to work from, or at least none that we understand as intelligent.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 19, 2008 22:26 UTC (Sat) by mmarq (guest, #2332) [Link] "" My personal opinion for the "next big thing" is self programming limited AI (not self-aware) computers that can figure out how to "do what I mean" when given vague commands in English. Maybe they'll ask questions to clarify the intent, or perhaps they'll pick the top five possible tasks and do them all, then provide a selection of results. "" Self programing !?.. not really happening. No, not that what you say is wrong, only that "selection of results" will always end up very short of satisfactory. AI is a simulation and selection, simpler tasks put together and complicated algorithms to do the selection and or simulation of a choice, but in the end is not intelligence because a machine have no choice and must always do what they are programmed to do... can never **REALLY** break main programs and still function. A machine can never have real intelligence, and be *REAL* creative because that implies the *freedom of choice* and a soul from where the energie of creativity flows... and machines have neither. The problem is the choice, and the true "CONSCIOUS" freedom of choice even in face of the worst possible tyrannies, which is the "must have" of creativity energies and indicative of the presence of a soul ... which machines don't have, and which can never be really programmed with any kind of not yet invented software.. simulations can be very complex indeed, but the Matrix and the Terminators thought very entertaining they will never happen. 100% guarantied. But a Darth Vader is possible, a Human soul in a majority mechanical body. "" After that is direct neural interface also with the self programming; probably an almost neural net design with many evolving programs running and the user picking the best results, providing feedback to the computer system so that it can keep the working parts and evolve the others. "" a machine reading minds, or reading elaborated thoughts and complex abstract ideas out of nerve electrical inputs ??... not happening, ever, its needed another mind and machines don't have it. But even if possible, for the sake of argument, would you submit to that ??... And what if your self programming does a lot of bugs, and your bionic arm starts to smash peoples heads, like melons, instead of shaking hands ??... A darth vader is possible but not with self programming software or something like that, that would have tremendous chances, almost 100%, of breaking up and mess up all neural inputs... an extremely dangerous uselessness... another example, thought robots are made of metal and much harder than flesh, they will never be martial artists... ever... its needed that kind of instinct/intuitive action that machines can never have. ( we should stop worry about robot armys... and worry about their loony proponents) The common mistake of AI theorists is not that simpler interaction man machine is not possible, but that a full interaction is not possible for a human level, not only because machines don't have responsibility because they don't have real free choice, but because all those actions of higher level are not body functions( which is another kind of natural complex machine) but mind-soul functions that are formless, pure energetic in a sense... belonging to a pure subjective world... not a objective one. So they are impossible to simulate. How would you AI simulate *creativity* in a machine, reason, intuition, instinct , sense of responsibility, LOVE,... other emotions... in other words, REAL INTELLIGENT CONSCIOUSNESS ??
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 19, 2008 23:21 UTC (Sat) by nix (subscriber, #2304) [Link] I'll start paying attention to that as soon as you can define 'real intelligent consciousness' and prove that it is nonalgorithmic. Of course biological neural networks are self-modifying nonlinear systems, but nobody has shown any reason why a computational system built along the same principles couldn't have the same external effects (including intelligence), and nobody has shown that they're not computationally equivalent to Turing machines (modulo all that randomness, which is easily emulatable, it's just thermal noise).
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 4:42 UTC (Sun) by mmarq (guest, #2332) [Link] "" I'll start paying attention to that as soon as you can define 'real intelligent consciousness' and prove that it is nonalgorithmic. "" Conscience is not information. It was already defined there, i think, in a more or less reasonable way: 'mind-soul functions that are formless, pure energetic in a sense...belonging to a pure subjective world... not a objective one. So they are impossible to simulate - in a mechanical way (i add) ' (of course there will be attempts, so what is impossible is not the attempts but call the results conscience ) 'How would you AI simulate *creativity* in a machine, reason, intuition, instinct , sense of responsibility, LOVE,... other emotions... in other words, REAL INTELLIGENT CONSCIOUSNESS ?? - capable of trying to understand the Universe (i add) ' All that are *real intelligent consciousness*, knowing is not a repeatable act of information, real knowing is up to the most intricate *whys* of a piece of information or knowledge... real knowing is discovering the ontological *whys*, not the factual information. Machines CAN'T DO PHILOSOPHY and or SCIENCE because machines can't think, they can only be reactive upon inputs and programs, repeaters and followers( LIKE OUR MASTERS WANT US TO BECAME)... not conscious... machines do not develop in an entirely SELF RESPONSIBLE, non programmable, true free choice, *the desires* and *generate the thoughts* upon truly deep subjects of contemplation and meditation... So here is a good definition of the functions of the SELF, the I, which machines can't have, neither which is possible to be programmable in any kind of software or machine, because is not mechanical but a complete transcendental think,... formless,... deeply with-in,... not *objectable* in any way or form outside of a human being... this SELF or I is the fountain of consciousness... And this SELF or I, is only able to be researched upon its PURELY SUBJECTIVE WHYS (actions and attributes), that is, another I or SELF is the only think that can really analise an I or SELF... or **JOIN in understanding with**... And machines are incapable of subjective actions even less subjective whys or programs... its never cold or warm but an interval of temperature, its ALWAYS true or false never *maybe* or something *sub-understood*... hesitate to go to some line of program, or jump to another... or program triggered auto(not self) create a totally different one... re-analise(not repent) and go back... or stop... maybe it can be somehow simulated, but the results will always be very crude and primitive ones, for the expectations around IA ... *I*(not programmed) must add... BUT... BUT... the from with-in reasons why(ontological), the self consciousness, the contemplative thinking way;...**I DARE YOU TO PROVE THAT IS POSSIBLE TO BE PROGRAMMED**. A little adding: ---------------------- We (humans) are the ALPHA and the OMEGA my friend...the only possible MANIFESTED GOD there is, is ALL of US(HUMANS) in all possible places and forms in the *ALL UNIVERSE* when understood like a *SUPERORGANISM*- **LIKE ONE**- That is the principal thing that all priesthoods of Power Hungry inclinations( our masters) do not want *you* (sub-understood the masses also) to know and or really understand, they will kill billions to prevent it if must... they and for their power sake, prefer **Like Machines** humans... with consciousness paralyzed by regret, guilt and fear... because there never was an original sin, but an original divinity... ...and they invent storys, if they do!!!?,... one must truly and deeply stand in awe!!!, with the amount of false promessies, and exaggerated claims, or outright BS signs and augments, that have enslaved so many and kill(still do) even more, like no other force on earth. So maybe IA will have tremendous amounts of funding for research never seen before.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 11:32 UTC (Sun) by nix (subscriber, #2304) [Link] Conscience is not information. It was already defined there, i think, in a more or less reasonable way:You have no evidence for this. (Well, OK, maybe you do have evidence. Maybe you've solved several problems in philosophy that are many millennia old. But I really, really doubt it.) In fact not only do you have no evidence for this, but your very definition is so vague as to be entirely meaningless, and abuses those words with clear definitions that it does use. (Go on: draw some predictions from it! Oops, you can't. Provide evidence for the existence of a 'pure subjective world'. Oops, you can't. What's a 'mind-soul function'? A function is a mapping from input to output within some domain and range. What's the domain and range of your 'mind-soul functions'?) Cartesian dualism is so passé. Machines CAN'T DO PHILOSOPHY and or SCIENCE because machines can't thinkWrong. Since 1976 argument has raged in the mathematics community over whether the enormously lengthy computer-generated proof of the four-colour map theorem is valid, even if it's too long for humans to understand: the general consensus now is that it is. There are lots of proofs like this. Computers can do mathematics and generate genuinely new results. Similar progress in other sciences is hard to see, but in a sense one could say that computers and people are partners in doing genetics these days: each would be pretty much helpless without the other (the only difference being that the biologists without computers can get replacement computers, but not vice versa). true free choiceYou keep on going on about this. If you have proved that this exists, you've solved another ancient philosophical problem. (Current consensus, inasmuch as there is any, is that it doesn't exist.) Prove that your free will is not an illusion. Go on. I'm waiting. ;) So here is a good definition of the functions of the SELF, the I, which machines can't have, neither which is possible to be programmable in any kind of software or machine, because is not mechanical but a complete transcendental think,... formless,... deeply with-in,... not *objectable* in any way or form outside of a human being... this SELF or I is the fountain of consciousness...Look, have fun with mysticism and word games, but don't pretend they're anything more. I've seen transcranial magnetic stimulation devices shutting down parts of people's brains temporarily and watched them change. I don't think there's a ghost in that machine, and until you can prove it's there you can't argue that AIs can't exist because they don't have one either. (A leading current hypothesis, backed by significant neurological evidence, is that self-awareness is as illusory as free will: that all you actually get is edited reports of what you already decided to do.) (Even if you're right, unless you think this 'complete transcendental think' originates from a deity, in which case you're talking religion, not philosophy, why on earth can't we synthesise one? It's not as though the formation of a brain is magical. We know a lot about it and will know more.) BUT... BUT... the from with-in reasons why(ontological), the self consciousness, the contemplative thinking way;...**I DARE YOU TO PROVE THAT IS POSSIBLE TO BE PROGRAMMED**.The extraordinary claim is that it is not programmable, as every other phenomenon we know of can be simulated in a computer, given sufficient power (the only ones you can't simulate accurately are truly random quantum phenomena and noncomputable phenomena, and we have no evidence whatsoever that the latter can be computed by people either. Besides, do you really want to lay your ineffable whatsis at the door of computational intractability?) So it is your claim to prove. We (humans) are the ALPHA and the OMEGA my friend...the only possible MANIFESTED GOD there is, is ALL of US(HUMANS) in all possible places and forms in the *ALL UNIVERSE* when understood like a *SUPERORGANISM*- **LIKE ONE**-This is formless mystic religion, not philosophy or science. As far as I'm concerned you've just conceded this little debate. You don't have any evidence at all, do you? That is the principal thing that all priesthoods of Power Hungry inclinations( our masters) do not want *you* (sub-understood the masses also) to know and or really understand, they will kill billions to prevent it if must... they and for their power sake, prefer **Like Machines** humans... with consciousness paralyzed by regretYeah, right. The authorities are engaging in a massive conspiracy to prevent AI. That would be more plausible if the Japanese government had not spent enormous sums in the 1980s on a big push to produce AI by the end of the decade: if government-funded CS labs and grant-funded organizations and corporations with funding from government and elsewhere weren't still working on things that may lead to AI without the least sign of any government protesting. On the contrary, they support it. I see no sign of any government trying to 'kill billions' to stop AI. (What is it with all the conspiracy theory nuts on LWN these days, anyway?)
All this reminds me of the old limerick Posted Jul 20, 2008 20:03 UTC (Sun) by sdalley (subscriber, #18550) [Link] There was a faith healer from Deal Who said "Although pain isn't real If I sit on a pin And puncture my skin I dislike what I fancy I feel". "Proving" that subjective experience isn't "real" does not seem a very useful thing to be trying to do, if it is even inherently possible.
All this reminds me of the old limerick Posted Jul 20, 2008 22:27 UTC (Sun) by nix (subscriber, #2304) [Link] He's asserting that Cartesian dualism is true and that (as near as I can interpret it) all men are gods. I think this is an extreme enough position to require some proof before anyone should be expected to believe it. (And yes, proving any of these things one way or the other is hideously difficult and perhaps impossible, which is why I was expressing such surprise at the degree of certainty mmarq was displaying.)
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 5:34 UTC (Sun) by klbrun (subscriber, #45083) [Link] I agree that a computational system designed as a self-modifying nonlinear system will show the same external effects (that is, intelligence) as biological neural systems. We won't build such systems because we won't be able to debug them. And we certainly wouldn't trust them.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 6:18 UTC (Sun) by mmarq (guest, #2332) [Link] See... Bottom line is, because those systems will not be really intelligent but complex and easily breakables ... ... and even unlike a solar system that although their enormous size operate like a clock for millions or billions of years, and so seem intelligent in a way, but are very simple to understand (Kepler is still plenty used, basically is the center of astronomy and space voyage)... ... so in a sense complexity is not intelligence, and solar systems will always be more intelligent (in a sense yes - but not conscious in any way neither) then any complex *simulation* machine system that man would be able to build. And this *solar system* intelligence that defies all numbers of chances or random, and have cycles, like it is exercising a choice or purpose, has nothing to do with self-modifying nonlinear systems.. quite the opposite. Intelligence has nothing to do with self-modifying nonlinear systems or neural networks... that is some of the external effects upon the interaction of something superior with MEMORY in human brains that AI research try to simulate... BUT MEMORY IS NOT INTELLIGENCE.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 11:42 UTC (Sun) by nix (subscriber, #2304) [Link] The human brain is 'easily breakable'? In a physical sense perhaps, if you get through the skull, but otherwise it works for billions of seconds with relatively few substantial failures, and most failures we see are markable up to problems with the neurological substrate (strokes, schizophrenia et al) or problems during formation (autism, epilepsy). There are other failures, but they only affect high-level stuff (mostly recently selected for and thus not as robust as the rest). You don't see the billion-plus neurons in the visual cortex go wonky for any reason other than biological problems, or exposure to stimuli that they weren't selected for and that were explicitly chosen to break them (optical illusions). I don't believe the solar system is intelligent (I can't tell if you're arguing for or against that), and I doubt (given your other post lapsing into muddled mysticism) that you have any evidence whatsoever for your oh-so-confident assertion that 'something superior' is involved in the generation of intelligence. (And your claim that MEMORY IS NOT INTELLIGENCE only proves to me that you haven't been paying attention to neurology. We know quite a bit about memory storage now, at least in mice, and a lot of what we'd consider 'intelligence' on a mouse level *does* appear to be handled in part by means of clever storage and nifty representations.)
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 22:24 UTC (Sun) by njs (guest, #40338) [Link] >or exposure to stimuli that they weren't selected for and that were explicitly chosen to break them (optical illusions). Actually, even this isn't a great example -- the problem with most optical illusions isn't that they fall outside the natural range of stimuli and so the system just shrugs and makes something up... it's that there are inherent ambiguities in visual perception (you're trying to reconstruct a 3-d world from a 2-d sensor array, now apply pigeon-hole principle...), and so the fundamental operation the visual system is designed for is "make the best guess". Visual illusions tend to be setups where even a perfectly designed, optimal in every way system will break down, because given the limited information available to it, the false/illusory percept *is* the best interpretation.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 11:35 UTC (Sun) by nix (subscriber, #2304) [Link] GAs are already being used to design things in actual real-world use. The results are generally highly nonlinear masses of feedback loops and generally impossible to understand, but they *work*. (At least, they work as long as you don't put them in conditions they weren't selected for.)
The Richard Hipp article was equally good Posted Jul 18, 2008 18:30 UTC (Fri) by smitty_one_each (subscriber, #28989) [Link]
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 18, 2008 18:42 UTC (Fri) by flammon (guest, #807) [Link] LT: 'I have a hard time really seeing what the heck Ballmer is doing. First the monkey dance, then the chair throwing. At some point he called Linux 'un-American', apparently because he doesn't like the competition. Then the cancer thing. And now this fixation with Yahoo! When will it end? I have a strong feeling that Linus reads Slashdot once in a while.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 19, 2008 23:53 UTC (Sat) by mmarq (guest, #2332) [Link] "" And when it comes to distributions, ease of installation has actually been one of my main issues - I'm a technical person, but I have a very specific area of interest, and I don't want to fight the rest. So the only distributions I have actively avoided are the ones that are known to be "overly technical" - like the ones that encourage you to compile your own programs etc. "" Why can't... even a good script in a package compile from source, and using good compiler settings for that matter ??... what else is there needed, instead of having a binary ??... the difference is little since Fedora RPMs are upfront nothing but a script. So why is that that script can't use *Source Code* instead ?? Ease of use as to do with ** GRAPHICAL INTERFACES **,... not if a package is mainly an text open source file, or a binary file. Most end users don't care!... most end users don't even have to start, the curiosity to understand the difference. So a graphical interface that compile from source upon installation, would be cazillions of times more easy of use then an apt-get command anything to most users used to windows. IMHO installing from source code files is what is missing from the Linux camp badly. Its disappointing to see Linus Trovalds fail to see that so obvious point. Compiling from source with a nice a neath graphical interface could add some practical meaning to the otherwise Open Source vs Freedom sterile debate... "" So what is freedom to you? Is it anarchy - the freedom to do anything you damn well want to do? If so, the BSD license is certainly much more free than the GPL is. Or is it any number of other ways to describe what "freedom" might mean? Often in very emotional terms, to boot? I'm not really interested in that kind of discussion. It's what I call "mental masturbation", when you engage is some pointless intellectual exercise that has no possible meaning. So when I try to explain my choice of license, I use the term Open Source, and try to explain my choice of the GPLv2 not in terms of freedom, but in terms of how I want people to be able to improve on the source code - "" But sorry Linus!.. still like the politicians... making speeches to each other. Not real difference from FSF. End users, most of them on desktop, neither understand what is the meaning of Open Source... or read the GPL license(any) for that matter... matter of fact they might have a much better grasp of what is *Freedom*... Now compiling from source, but promoting HEAVILY GRAPHICAL INTERFACES, might do a tremendous job at imposing a view of pragmatism... Unless RH don't like desktop and graphical interfaces, and a better way to keep installation and configuration of systems out of hands that don't understand the CLI and some basics... no RH Entreprise for Dummies... though with GUIs, lots of people could remember a lot of dialog boxes/buttons positions!!...
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 0:40 UTC (Sun) by vonbrand (subscriber, #4458) [Link] The point of distributions is being able for others to find (and fix!) the problems that don't interest you. Plus having to wait for my machine to do what some other machine could very well have done for me is puzzling, to say the truth. Yes, I used to build much of my stuff myself a long while back. It made sense when we had Suns around (and had to get an oldish gcc to build a new one, and build a decent X, and needed bash, emacs, (La)TeX and assorted other stuff), and I continued doing so for a while afterwards when I got Linux. I soon found out that you need to focus on what you work on, and get a solid workbench for your primary work. In my experience, you outgrow "building from scratch". Just like you outgrow the "one distribution a week" phase.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 1:49 UTC (Sun) by dlang (✭ supporter ✭, #313) [Link] it's also a matter of what the return on the investment (of time) will be. currently the ditros are producing binaries that work well on current hardware. there was a time when they were producing i386/i486 binaries that were relativly inefficiant on the then-current hardware (around the time that processors were hitting 1GHz I was seeing 30% performance gains from custom compiling a kernel for my specific CPU and hardware) as hardware gets faster the sight gains from custom compiled code are much smaller then the losses caused by poor/inefficient design, and both are normally swamped by the new hardware. I tinker occasionally with source-based distros, but it's not a matter of efficiency or performance that draws me to them, it's the ability to avoid a lot of the dependancies that current distros drag along with them
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 20, 2008 5:43 UTC (Sun) by mmarq (guest, #2332) [Link] See!! That is the reason why there is a cliff between *masses adoption*(and this means desktop inexorably) and potential economic grow that will enable more and more people to take a deserved income or reward out of the all FOSS enterprise. FOSS participants only think in terms of developers and IT professionals ways... not in terms of the masses of end users. When will be the Desktop Year ??...a few seconds after the main FOSS developers and IT pros involved, start to think the other way around. Is it reconcilable ?... BY ALL MEANS... that is why GUIs were invented,.. and why VOICE, and maybe sometime GESTURES in front of a camera, will be next... Extensive GUI everywhere, and so a more centralized effort around desktop, is by my understanding the real ease of use way... and my point of a very good package management GUI program, that can do source code compiling and installing by point and click, is by all measures a ease of use feature... MUCH MORE if the packages are binary but the installing program is a CLI. WHY ?... it is much more easy to remember a pretty graphic then a combination of words... more so if those are not real words but somehow more or less cryptic commands.
Linus Torvalds, Geek of the Week (simple-talk) Posted Jul 21, 2008 4:03 UTC (Mon) by drag (subscriber, #31333) [Link] > Why can't... even a good script in a package compile from source, and using good compiler settings for that matter ??... what else is there needed, instead of having a binary ??... the difference is little since Fedora RPMs are upfront nothing but a script. So why is that that script can't use *Source Code* instead ?? It can. The problem is that you need to have a fast computer, use up lots of disk space, and use up a lot of the user's time. A user has to have all the development tools installed. They need all the dependency's 'dev' packages installed. And of course those are all installed by package's also, so they get compiled from source also. (and all those dependancie's dependency's dev packages etc. etc. etc.) So instead of a Linux install taking up 500-2000 megs of disk space your using up 8000. Instead of a 5 meg package taking 30 seconds to install your downloading a half dozen and taking a half hour to two hours to install a piece of software. Seriously. Copying a file is fast and efficient. Compiling is slow and irritating. For a given architecture there is only the need for one person to compile a program, for everybody else they can just copy a few small binaries. This save times and eliminates variables. Saving time is good. Saving electricity is good. Saving frustration is good. ---------------------------------------- Seriously. Binary distribution is the way to go. It's the only way to go. The problem is that we have all these different distributions. Even Ubuntu, which each release is a snapshot of Debian Sid + Ubuntu's special sauce, doesn't bother to keep package compatibility with Debian and other Debian derivatives? Why? Laziness is all I can see (and the resulting time saving and all that). Making it easy for other people is difficult for yourself. There is probably still a justifiable purpose to having multiple distributions, but in reality there is absolutely nothing that I can do with Redhat or Fedora or Debian or Ubuntu that I can't do with any of them. They are all using the same software, same kernel, same source code. There are a few variations and changes in the default installs and a few scripts and configuration incompatibilities... but who really cares at the end of the day? I vastly prefer the Debian way of doing /etc/network/interfaces and their ifup and ifdown scripts and all of that stuff. It just makes a lot of sense, it's easy to understand what is going on and it's easy to edit. In comparison Fedora's network stuff is a nightmare. But so what? It's nothing that I can't handle and lots of people use it all the time so at least it has to work. That's good enough for me. It's easier to learn one of them instead of both of them if I have to work with both Fedora and Debian systems (which I do). I think that the days of 'competing' distributions is over. I really needs to die. I have no problem with one distro focusing on improving one part of the 'linux system' and trying to use that to put themselves ahead.. but the basic 'gnu' or userland is a done deal. It's been done to death and it's possible to get a basic OS that can be used by everybody for everything. This includes the package management system. Developers _should_ be able to distribute binaries with installers and it _should_ work with all 'mainstream' distributions and integrate with their package management system. It should of stopped being a big deal 7 years ago. This is a major problem and it needs to be addressed because it's simply a core usability problem. Instead of re-inventing the wheel a dozen times, why don't we start figuring out how to make it fast, easy, and efficient?
Distributions and Source Compilation Posted Jul 24, 2008 20:26 UTC (Thu) by zlynx (subscriber, #2285) [Link] Source distros like Gentoo provide an immensely valuable service to the world. Without Gentoo, many developers and their projects would go about life believing there are only two versions of GLib, or that glibc is the only Linux C library, or that it is OK to do text comparisons of library version strings in their ./configure scripts (this breaks when comparing version 9 against version 10, you see). They also get early notice of how their C++ code is non-standard and is broken when compiled with GCC 4.4 alpha. Because of Gentoo and adventurous users like myself, projects get these bug reports (and often patches) which force them to remain flexible and get things done *correctly*. This avoids the situation Windows has developed, where projects can only be built from source with the precise versions of Visual Studio and companion libraries that existed in 1997. Gentoo was also *the* best distro for early AMD-64 users. It took Debian and Redhat months to catch up. Now the reason there are so many distributions. Well, take Debian and Fedora for example. They operate in different ways. Fedora is more able to take off in new directions, such as converting everything to PulseAudio and DRI/DRM version 2. Debian has more developers with their own opinions, which makes it very hard for Debian to take off in a new, risky direction because of all that inertia. Or what if some people want to build a Linux that boots to the desktop in under 10 seconds. Are they going to be able to convince all Debian developers to build against uclibc, rewrite init scripts into C compiled submodules of a new custom "quickinit" and build a no-modules Linux kernel without SCSI drivers? I don't think so. Thus, distributions.
|
|||||||||||||