Alan Kay - Programming and Scaling

Alan Kay – Programming and Scaling



Views:13958|Rating:4.83|View Time:56:39Minutes|Likes:197|Dislikes:7
For cool updated about the future of computing, follow me at

This video is about Alan Kay – Programming and Scaling

so of course we never want to spend too much time looking back especially when talking to students because the world is in front of you but one point I wanted to make and this has always been true and has never reflected adequately in any rewards that a person might receive and that is that when good work gets done it almost always gets done first in a powerful research community of very friendly and very critical colleagues who know how to argue the right way so as you're looking ahead for yourself finding these people in general is critical and finding them in particular is critical so these are just some of the people that have been closely associated with some of the work we've been doing it viewpoints Research Institute and very few awards actually are given to groups that this is a real mistake in our field because I believe in our field that you not only have to have ideas you have to get them implemented and it is often the case that people with ideas are not so good at implementing and so you need to have a kind of a team of different kinds of talents who are able to work together one little look-back though it could be so I as we as the term is meant today I made up the term object-oriented programming but I made it up for something different than the term stands for today and what the term stands for today this could be the 50th anniversary of it so that's quite surprising to think it might go all the way back to nineteen 61 and it goes back to a very surprising form which is actually a machine so there's no object-oriented programming language in 1961 that I'm aware of I actually started in computing myself in 1961 so there was nothing like an object-oriented programming language back then but there is a piece of hardware that can most easily be explained to you is that this was a computer hardware that executed a set of instructions that look like the byte coded VMs that you're familiar with from Java in fact those byte coded VMs that Java have came originally because they were the basic instruction set on this machine so this is one of the first attempts at a building a higher-level machine to reflect the idea in the early 60s that we were going to program in levels of language higher than C and Bob Barton here invented this machine and sadly the hardware of the day is not as advanced as this machine was so for those of you who think we've been engaged in steady progress in some form of Darwinian processes that are making things better for us it's actually quite the opposite quite the opposite in history you can find really incredible geniuses like this man who anticipated so strongly that we can only emulate his ideas and software today intel hasn't gotten around to understanding them yet so this I find really quite sad on the other hand it gives you a sense of this person and if you want to read an interesting paper probably the shortest greatest paper ever written in computer science you can read this one written in 1961 by Bob Barton new approach to the functional design of a digital computer it's only six pages long and it lays out what a felicitous execution environment for higher-level language would be like and how it would be realized in a computer this computer was built by the way it lived on longest in Germany the German banking system loved it because it was uncrashable why was it uncrashable because it was the first hardware or software system to have what we call capabilities today an application programmer simply could not write a program that could hurt any other program on the computer that was built directly into the into the hardware of this machine and the reason I'm showing you this just because part of it is everybody should understand the history of our field if we were physicists and we failed to understand what Newton did we should be shot and but in fact in computing we are very happy not to understand what most of the founders of our field actually did so I always like to pick one of them in the generation before me and explain what he did but reason I brought it up here is that I had a course from this guy this guy was took a leave of absence from borrows to come teach at the graduate school I was at and in this course which was about advanced systems design instead of teaching this course he handed out a list of all the papers we should read he said there are a few things known about advanced systems design in 1966 they're written down in these papers and I expect you to read every single one of these papers and understand them and then he said but it is my job to firmly disabuse you of any fondly held ideas you might have brought into this classroom and translated into less flowery English what he proposed to do is to take all of our beliefs about computing and destroy them and that is what this class was the best class I've ever had in a university or graduate school and why was this a good idea well we'll see in a second that the biggest problem we have as human beings is that we confuse our beliefs with reality so we seem to live in a reality here but actually it's a construction that we look at one way and people from another culture might look at it another way and my little dog Watson might look at this whole thing from a completely different point of view and when we treat a construction as reality you are actually binding the kinds of thoughts that we can actually have and worse than that we're picking a tiny part of the past that led to this particular construction as the only past we ever drawn so what this guy did is just demolished everything that we thought including things that he actually believed in so he demolished everything no no proposition was safe and at the end of it we actually were free so we could then learn computing from the point of view of it being simply opinions and a little bit of practice that have been done so far up that way but might not have anything particular to do with a desirable future so here's a paradigm that existed back then and is pretty much what we do today and I call it the make and fix paradigm and it started off with data structures and procedures they were like tinker toys maybe a little bit like clocks we fit we make gears we fit them together called api's they have to fit pretty closely for the thing to work and that's pretty much the only architecture we have and of course you can make a tic-tac-toe computer out of tinker toy this is Danny Hillis has a few thousand tinker toy in it but like such things was once dismantled to be shipped to a museum and when they put it back together and never worked again and the reason was it was just luck that all of these Tinkertoy actually work because there was play there there strings inside the thing and the strings weren't the same length anymore and that reminds you of any programs and this is this murky thing is actually Mexico City has anybody ever been to Mexico City so it's 20 million people couldn't be more smoggy and it's kind of a mess and it's just the kind of it's just a slightly more extreme version of a lot of American cities where individuals have carpenters have decided to do a house here or a house there and the problem is is that each individual little ant doing something has led to a totality that is almost unbearable and yet 20 million people are forced to live in it and to me this is what software today is kind of like it's basically using almost the exact same architectures as we're known 50 years ago because even the so called object-oriented languages today have lots of setters in the objects and when you have a setter in an object you've turned an object back into a data structure and so you get this huge amount of scaling of complications but you don't get a lot of bang for the buck it's not clear what to do with it but what yeah so if we look at software in terms of lines of code for instance a one book of software is about 20,000 lines of code and a foot of books is about I have it here a meter is about a million lines of code so if you stack up a meters height high a book it's about only a million lines of code okay so if we look at some software systems for instance Vista has 120 million lines of code in it so that's a hundred and twenty meters of books to read and six thousand books and another 140 million for the apps another seven thousand books so we're talking thirteen thousand books for the simplest part of personal computing that we're all used to anybody here read thirteen thousand books anybody here read one book okay just checking so there's a lot and I know a big software company whose name you might have heard that has to deal with a live base of about 350 million lines of code and as they say about 145,000 screens so screens is a term that was used a long time ago for 3270 terminals and so if they call their user interface today made out of screens it's probably that they just adapted those old 3270 screens so this is about seventeen thousand five hundred books worth of software nobody in that company knows what it is but they have thousands and thousands of people maintaining it so the question we should be asking as computer scientists first let alone engineers there's computer scientists we should be asking is that how much software we have to write to get that much functionality or is something wrong so one way of looking at it is if your are two curves so the complication curve here I just put up you know about a a factor of a hundred this and so we could see that while we like the complication curve to run at the same rate as the complexity curve it actually is way ahead of the game so for a certain level of complexity like say in personal computing the actual complication might be a factor of a hundred or a factor of a thousand like all but one tenth of one percent in the Microsoft software suite might be useless code code that is there perhaps for compatibility reasons code that they don't dare take out in fact I know that for example if everybody here is probably used Microsoft Word and it's quite surprising that Microsoft Word when you have paragraph justification on actually has selection errors to this day if you ever had that happen got justification on you click in there and you start typing and the type character goes somewhere else that bug has been there for more than twenty five years it's been reported literally thirty or forty thousand times and they cannot find the bug this is the most elementary thing that you ever do in a word processor is to be able to select where you want the character go they can't find it and fact I've heard that part of Microsoft words listing has actually been lost they've had to recreate it from from assembly code so that's kind of a state of affairs that we probably should be thinking about because when you start getting into factors of a hundred and factors of a thousand and factors of 10,000 we're actually talking about the difference between somebody being able to comprehend something and maybe not be able to comprehend it at all so if we go back to our diagram here we can pick up some issues that we might care about now say something about the communication with aliens issue a little bit later what that is and if we look in this morass of stuff there's one interesting thing that isn't like this that is huge and yet it is tidy its level of complication exactly balances its level of complexity and that's the Internet so it's worthwhile thinking about that that even in a poorly done C program the Internet is actually controlled by less than twenty thousand lines of code itself and that twenty thousand lines of code can be written in about a thousand lines of code so tcp/ip is a kind of a universal DNA for an architecture of billions of nodes in a dynamic system that has never been stopped has never been stopped from when it first got turned on in September 1969 so replaced all of its atoms at least once and has never been stopped it has been changed it has been improved but in fact it's an eternal system it is scaled well and so forth and I'm not talking about the web of course web is a mess but the internet was done by experts you had a couple of those guys Vint Cerf and Bob Kahn here at different times these guys knew what they were doing and boy when you look at this I think why it is why did the internet turn out so well and why can't we do regular software that way because you could imagine taking the inner as a model for doing software modules why don't people do it and it's the way I look at it is a little jump of qualitative kind happen here and that jump is kind of a little jump from gear type things to biological type things and in fact those of us who had been biologists contribute some pilot biological ideas to the design of the Internet because of course the Internet is one of the few human artifacts that we've made that behaves like about like something that's alive in that it repairs itself while it is still running it is able to expand and grow and so forth and so we can think of the Internet is in it was made but it has some of these biological ideas in it so some of the same kinds of ideas for scaling were done in software also around the same time and some of those were successful at Xerox PARC but one of the problems of these some of these semi new ideas here is when they got translated into sea based forms in order to be recognizable to people the things that were like this old paradigm were emphasized here so for example you got this really terrible idea of having a so-called object-oriented system and turning it back into data structures by putting setters in which allow the very same disastrous remote assignment statements to to be done and turns the system into a much more fragile way of thinking about things but I think that what's actually going on here is that there's a new way of thinking about things that represents a brighter future for programming and if this is makin fix for instance you can fix a clock but you have to negotiate with the system so at some point the systems are too complicated for a human being to go into too many feedback loops too many self regulatory systems and stuff and you wind up at best negotiating with the system now the stuff we're doing in here so far is in between we built it by hand but we made a system as we've made up many other systems in software but we haven't made the systems capable enough to be able to be negotiated with so I just want to have you think of it in your head of what it would be like if you're doing a software system in some new form of development interactive development kit what would that software system actually do besides just doing what you asked it to do could it for instance participate in readapting the software for future needs about 80% of the cost of all software that is deployed in industry comes after the software has been successfully put out the first time so the life cycle changes that are required of the software and the integration and reintegration actually dominate all the costs of software but because of the way companies have to report their earnings and their expenses companies cannot charge upfront for all of the stuff that you really need to do to software to allow it to be changed downstream so what they tend to pay for is just enough money to make the thing run to its original specs and hardly any money to be able to fix it up later because they just amortize those costs of the future and the costs are enormous so if we take the internet as one area and then a slightly more advanced way of thinking about things is that once we start thinking in biological terms we also are thinking about Ecology's and I hope many of you have thought enough about the Internet to realize that there are levels of communication protocols that could be put on top of tcp/ip for example interoperation protocols that could eliminate the dependence on specific operating systems and I hope to get to that in a few minutes okay way down the road what is actually being put together is something that is going to have to be discussed in psychological or mental terms because the if this is manipulation up there just wanted to make a little comment about biology because sometimes if we study computers we don't learn other things so we can I'm going to read this diagram both way so at the inch scale now there's what twenty five thousand Mike twenty five thousand four hundred microns per inch and a ecoli is about two microns long so it's about 13,000 microns per inch a little over ten thousand per per centimeter and so if we blow it up by eighty thousand we get to see this and of course it's very active so we wouldn't see this we would see a blur and if we take the next section up there we get something that looks like this where we see a bunch of large molecules we take a section of that so we're now a magnification of about 10 million we see a medium-sized molecule has about five thousand atoms in it and these are actually water atoms so if you like books like this it's called the machinery light of life by David Goodsell this is a molecular biologist who likes to paint and he realized that doing watercolors of what is actually going on there could be more clear than computer graphics or anyway this is a marvelous book for getting an idea so these are individual water molecules these are calcium ions 5,000 atoms or so and if you see the size of this thing compared to it in there and see it in the size of the larger thing they're about 120 million molecules that are not water in a single living thing so one of the things to think about in computer terms is the scale jump from the next lower level of architecture which is polymers like proteins chains of atoms to make long molecules the scale jump to what what it needs to do to make a living thing is incredible compared to the scale jump we do when we try to make objects and I think one of the mistakes we made years ago was to make objects too small and we missed this idea that you might get a lot more out of making a much more capable more universal object and try to build things in terms of that and just for comparison there are just slightly fewer tinker toys in Danny Hillis tinker toy tic-tac-toe playing computer as there are in one of these proteins so so a protein you could imagine assembling by hand but you could not imagine assembling a bacterium by hand okay got aggro on okay so here's a piece of context that I always like to ask students about so here's an interesting thing about all of us is pretty much from the time we were born every person in this room was praised for being bright right we just talked up a storm and we were interested in this and we were so clever and we were encouraged most of the time to be clever it seems like a good thing and maybe it is because here we are but imagine being born with twice the intellectual capacity of Leonardo smarter than any of us no question twice the intellectual capacity of Leonardo but in 10,000 BC how far are you going to get how far did Leonardo get Leonardo could not invent a single engine for any of his vehicles think about that maybe the smartest person of his time but he's born in the wrong time his IQ could not transcend his time so IQ is really not worth that much because talking about vehicles here's Henry Ford nowhere near Leonardo but he happened to be born in the right century century in which people had already done a lot of work in making mechanical things and so it was Duck Soup to make a car that worked and sell millions of them for four hundred and seventy five dollars apiece and of course we have Daimler and Benz here in Germany so basically when you born at the right time a whole bunch of things are open to you so knowledge in many many cases just Trump's IQ I and this is because there are certain special people who invent new ways of looking at things so Ford was powerful because Isaac Newton changed the way new with the way Europe thought about things so he changed an entire way of things and it's one of the wonderful things about the way knowledge works if you can get a supreme genius to invent calculus those of us with more normal IQs can learn it so we're not shut out from what the genius does we just can't invent calculus by ourselves but once one of these guys turns things around the knowledge of the era changes completely this is why in the United States at least computer science is almost dead because of course I would get a lot of argument about that because there are many people there but the outlook there has gotten pretty weak and so the way I think of it is the outlook at you have on something is what is going to give you those extra IQ points but if you have a really powerful outlook the knowledge you'll gather will be powerful and then if you have some extra brights if you happen to be clever when you have those first couple of things then you can be very powerful so I think of knowledge as silver and outlook as gold and IQ is kind of a lead weight this is killing computer science because a lot of people in computing don't bother learning anything because it's not that difficult to learn how to program badly right and you can be clever and write a program in JavaScript to do lots of different kinds of things the problem is you there's no place to go with it so when I look at our field today from this 50-year perspective well looking back I would say the absolute number of four four and five Sigma IQs is actually has actually increased so if we were to take all of the people in our field vastly more and look at the superbright there's just more than there were 40 50 years ago but the relative number is much worse because the total number is increased tremendously and the absolute number of the superbright is not increased commensurately and so there's an enormous amount of noise from the middle today in the field an enormous number of artifacts from them similarly there's a lot more knowledge of computer science today than there was 40 50 years ago but the general knowledge of most of the practitioners has only increased a little bit the level the small size of this increase is almost staggering given what there is to know that has been written down in various places and then the number of outlooks that are powerful has gone way up over those years and but the general outlook is hardly more powerful than it was I'm talking about programming here this is quite static so something something is going on that reminds me a lot of pop music right there if you sit down and learn music there and look back over the last six centuries and it's just some of the most incredible talking just about classical music I'm talking about jazz any kind of develop music it just amazing the riches that are there yet pop music uses 1/100 of 1% of that and so I think what what's happened is computing has turned into a pop culture and the universities are not helping in general at least in the u.s. so Cicero anybody know a good quote about Cicero having to do with present and past let's check out your classical education here so you know who Cicero was he was one of those old Roman guys so Cicero once wrote he who knows only his own generation remains forever a child and basically what McLuhan is saying here is we tend to be trapped in a single conception of the present and it makes it hard for us to see it in any terms except the terms that led up to the present that we're in and he said and that means it's hard to conceive of a future that is an incremental on the present makes sense right the president is overwhelming to us and there are all these little things that we can do a little bit better on it that gives us a kind of a model that's like this so here's some rain falling outside making a little gully by random and it happens that a gully by random just happens to be more efficient at making the gully bigger so if you just let that go on for a bit you get these guys there's no reason for where these are just some random rain disturbing some things there but once it got set up it got interested in itself just like our brains do so this is not a bad model of the way human memory works we start learning something it was very hard for us to see what else is going on and so there's this famous picture of Conrad Lorentz who discovered that the first thing a duckling saw when it came out of the egg it thought was his mother and so if Lawrence arranged it so he was the first thing they saw so the little ducks would follow him around and when he went out swimming they'd follow him around and so I think that's the way people often approach well your first programming language right so probably the most disastrous thing you can ever learn is your first programming language even if it's a good programming language and the reason is is it's tends to become computing and so it might be a better idea to learn two or three programming languages at the same time even though that's a different kind of struggle but at least relativize what people think computing might be so we got this present it comes out of one set of things in the past that we're daily aware of it gives rise to an incremental future this is a lot like academic paper writing in the US I don't know whether you have to do it here but so almost all the papers written in in the US and computing academia are counted so the University of California counts all papers written by academics they don't care whether the papers are good or bad they don't care whether a professor has a really great idea every five years they do care if the professor writes 10 or 15 or 20 papers each year and the same with graduate students because graduate students are taught to follow in this mold and often the professor has his name on that very same paper the graduate student in increases his count so there are a lot of factors here but the truth is that the past is vast it's enormous there are billions of people contributing to the past and every time we think the present is real we actually cannot see the rest of the past so we have to destroy the present one of the ways of doing it is to try spend a little time each day trying to think of what you take as reality as being just a result of that rain falling on the ground and being intensified get rid of it once you get rid of it it's a scary situation because you've said I'm not going to have anything based on the course that's not possible you're just trying but sometimes you get a little feeling this is not an idea it's just a feeling it's like an odor of perfume but the fun thing is that little feeling can actually lead you to look in the past in different places then you would normally do and you can bring those up to that feeling and once you do that that feeling starts expanding into a vision and the vision expands into an actual idea so this is one way of doing it and some of the most creative people I know actually operate this way this is where these ideas come from that are not just incremental to the present they come out of vague even muscular sensations that you have to go chasing to try and find out what they are you try and get the idea too early it can only be in terms of the present what we do so everything that human humans make is technically technology so I'm going to and we have a different meaning for technology so I'm going to change the Greek root of technology is technique technique means whatever human make humans make don't let's call it the arts here and in the arts we have tinkering I'm going to do these in historical order so mammals are creatures who tinker especially primates especially us and engineering is a kind of principle tinkering so it came before math and science what does it mean by principles if you don't have math and science well if you build a building and it falls down you just build it with twice the cross-sections of all the beams you don't need any math and science for that if it doesn't fall down you write it down in a book and you wind up with this big cookbook that by the way is the way Babylonians did math the Greeks hated long books of things like that that led them to invent the company math we use today that came next science as we know it today came about 400 years ago so anybody here know how long our strain of humans has been on the planet how long have we been here this is it's kind of important isn't it for context so you know about so it happens that we inherit more from our mother than our Father because mitochondria and its DNA don't fit into sperm but they do fit into OVA and so when we got a fertilized egg we cannot only get the union of the two nuclear DNA s but we also get the female mitochondrial DNA and mitochondria are a little bacteria that were captured a long time ago to help with energy production in cells they were originally separate organisms so it turns out you can actually run computer simulations on the amount of change in the female line of humanity should do this sometime because it's kind of interesting and it turns out that if you run it back we were all defendi descended from one female now of course there are other females around back then but that female lived 192 thousand years ago so our particular strain of humanity has been here but for about 192 thousand years think about that we only invented science 400 years ago that's how dumb we are think about it that's why IQ doesn't count we never were smart enough to invent science we blundered around amusing ourselves with stories for hundreds of thousands of years so almost everybody in these fields winds up doing a combination of them and you kind of want to be in the sweet spot you want to do a little tinkering a little engineering a little math a little science and really the difference between the fields is more of temperament than the kinds of things that you learn some people are idealist some are pragmatists now if we look at computing it's mostly tinkering a little bit of engineering maybe a little math hardly any science and when we say engineering we don't mean real engineering real engineering uni Empire State Building in New York was built in less than a year by less than three thousand people including dem Demmel it demolishing the original site and occupancy so it's built at the rate of two stories a day the steel was still 110 degrees 40 degrees centigrade when it came from the factories in Pittsburgh to make the Empire State Building that's Engineering so whatever it is we do is like maybe what Egyptians did making pyramids piling things on top of each other but in computing we could not organize 3,000 people to do a major thing in less than a year right have you ever heard of it ever No so if you take the word engineering seriously and I do I don't mean it as a metaphor I'm saying engineering let's not do away with what engineering has meant the last three or four thousand years we take that word seriously we haven't got it yet in software we really don't know how to do it okay so this is my favorite hairball it's actually from a woolly mammoth so that's a big hairball I'm gonna get rid of it because but you can think of it as something complicated I'm going to replace it with stars because they're nicer to look at but if you think of we're given this phenomena that's the basic thing it could be a hairball it could be they're all I abstractly hairballs and what we want to do is make sense of it in science and we have language which is computer scientists you know can all be made out of the operation NAND or nor which actually invented by a guy by the name of Charles purse in the 19th century so we can make all things having to do with language out of that one operation and what we want to do is to come up with something in language that closely mimics mimics the phenomena that we're interested in and forget in fact because of the limitations of our brain that's about as far as we can go because the brain contains various kinds of languages that does not contain stars and so we're interested in the relationships between those so the reason I'm explaining this is because when I ask graduate students in the United States what is their meaning of computer science they always give me an engineering definition which implies that they have no idea what the word science actually means this is what it means but science doesn't care whether nature made it or whether humans made it so we can take a bridge do exactly the same thing make a t-shirt explains how this bridge works and what's cool about engineering is once you make that t-shirt you can make another bridge get those engineers to make more interesting bridge for a scientist to look at and you can take this rather far those are my favorite bridge in Japan longest suspension bridge in the world and to give you an idea of how big it is there's the Great Pyramid of Egypt and the Empire State Building next to it so it its towers are actually the size of the greatest building created in the early 30s this is some kind of bridge makes you feel good to be an engineer at least it makes me feel good because I this is you know the ability to really make engineering and science harmonize with each other there's no greater music so computing we've got computers we got software here's John McCarthy he can put the abstraction of all that stuff on a t-shirt that's the Lisp interpreter in itself one of my favorite things for more than 50 years ago 1958 if you don't have a little tear in your eye when you look at that you're missing something really really important about our field this is like one of the greatest things that has ever done because the same thing with the bridge once somebody does that man you can start inventing things that are much more you can invent computer things that are like like living things and make Internet's and so forth so this is really cool stuff and you can also look at personal computing and ask how many t-shirts we had an Empire State Building worth of code from Microsoft but we should ask how many t-shirts does it really take and so one of the things that we've done in our approach is finding that problem difficult it's really hard to think this thing through because geez we murse tin this stuff it looks large it's complicated so what you can do is make a little Frankenstein's monster of a system it isn't the one you actually want exactly but it's kind of cute and it's kind of scary and if it's also kind of small then you've actually made the thing that will allow you to think about this other guy and Olay enable you to make a baby step towards capturing millions of lines of code in a few tens of thousands of lines of code and so that's one of the things we've been doing so my website the website of viewpoints search Institute is Vpr i.org and there you will read more about this project and I'll just show you one example here from a bunch of principles one will be the math wins one tiny Model T t-shirt programming and a little bit of particles and fields so this is one of our colleagues dan amma lang and he was charged with the goal of doing all of the computer graphics that you see on a personal computer and so that's several millions of lines of code for instance we took Cairo as an example Cairo is the rendering system and composing system that's used by Firefox so it does most of the graphical tricks a few million lines of code and turns out dan has a mother and his mother as a high school geometry teacher and she asked him well what are you doing son a few years ago and he told her and they sat down didn't come back to work sat down between Thanksgiving and Christmas and worked out a new way of thinking about the involvement of arbitrary polygons with pixels in math and it looks like this so this is the formula they came up with so with that this is without any super sampling or anything else like that you just execute do what that formula says and it will give you the exact shading for perfect anti-aliasing so of course you can't run that formula so the next thing he did was to make up a language there was rather like this mathematical language but which was executable so you can think of this as a domain-specific language although it's fairly general and it has math in it it's stream-based it's highly parallel and so forth and these the 40 lines of math over here comes out to be about 80 lines of runnable code now you can run it and you get something like this so the rendering part will render high quality characters and simple shading and and so forth like that so so that's 80 lines of code and of course there's more features and Dan also has a brother and this brother is very good at math also so he and Dan one summer made up the 26 standard compositing rules that are used in Cairo and all of them together was about 90 lines of code all right because if you think about it if you don't worry about optimization or anything else like that the mathematics here in the right kind of language it just falls out you're actually writing down what a mathematician to write down is the actual relationships and you get each one of these things does a different compositing there's the rendering stuff sampling for making pictures pen stroking so the total today is a little bit under 400 because we added a few more features to it but basically all of the graphics that were used to in 2d on a personal computer can be done in under 400 lines of code okay that is a big difference compared to maybe one or two millions of lines of code in Cairo like you can look at all of it 400 lines of code is just a few pages well as a hitch and I'm going to end this talk with just telling you how we get written the hitch of course is you have to make this language run right so you don't want to spend three year you'll come up with the perfect language and then spend three years trying to get it run so you have to have another language and this is based out of a paper from 1964 about how to make the swirls simplest compiler compiler done by a guy at UCLA Val sharee and we did a modern version of that here's Alex Werth who did it and so the parser for making the graphics language is 130 lines of this language which is called Omata and the the optimizer and the compiler to a language like C is another seven hundred some-odd lines so that's what it takes to make the graphics language so it's a little bit larger but the nice thing about this compiler compiler language is you use it for each of the domain-specific languages that you're making things from and of course you have to make that language but what's nice about it it can make itself and it can make itself in only 76 lines of code and it can optimize itself in 30 lines of code so you get the idea here all right so the idea here is that each problem space has a particularly felicitous point of view point of view is worth a D IQ points if you find that point of view you're going to be able to write down the relationships more or less directly and that helps you think about the problem you're working on and then if it's easy to make that domain-specific language you just came up with run you're in business because all of a sudden you've got a system that's actually it's actually working so I'm gonna so we're out of time here and I think I'll leave you with that I was gonna I think that's more than enough information but this is I think one of the points here that's interesting is that this idea is not only not new today that's why they're called dsls today it's a term we have today turns out this idea was in full flower 50 years ago nobody knows about it today because the terms change it was they were called problem oriented languages 50 years ago POS and for a variety of reasons this whole technique of making software systems very quickly and very in small form got dropped when the commercialization of personal computing happened in the early 80s note it's not taught as a as a major technique and very few kids in the United States know that it's easy to make a compiler compiler language that is much more general and much easier to use than say yak so you learn yak and you're convinced that it is hard to make a language but in fact yak is bad the actual techniques of making a language or actually run much easier to do so let me leave you with this idea that whatever it is that we think we know today we should shut it off periodically and see if we can find other things to learn about thank you

6 Replies to “Alan Kay – Programming and Scaling”

  1. May 10th

    @10:20 Alan says that having setters in objects makes them a data structure. @19:20 He says putting setteres in allows the "very same disastrous remote assignments statements to be done, and turns the system into a much more fragile way of thinking about things." What does he mean and what is the better alternative that Smalltalk / real OO propose? Thanks for your help.

Leave a Reply

Your email address will not be published. Required fields are marked *