1 00:00:03,526 --> 00:00:04,268 Hello. 2 00:00:04,268 --> 00:00:07,714 In this video, we're going to talk about something that I've referred to as 3 00:00:07,714 --> 00:00:09,526 the economy of programming languages. 4 00:00:16,202 --> 00:00:18,524 So the idea behind this video is that, 5 00:00:18,524 --> 00:00:23,172 before we get into the detail of how languages are implemented or designed, 6 00:00:23,172 --> 00:00:27,601 I wanted to say something about how languages work in the real world, and 7 00:00:27,601 --> 00:00:31,130 why certain languages are used and others are not. 8 00:00:31,130 --> 00:00:35,520 If you look around, there's actually a few obvious questions that come up to anybody 9 00:00:35,520 --> 00:00:39,320 who thinks about programming languages for more than a few minutes. 10 00:00:39,320 --> 00:00:41,920 One question is, why are there so many of these things? 11 00:00:41,920 --> 00:00:47,050 We have hundreds, if not thousands, of programming languages in everyday use, 12 00:00:47,050 --> 00:00:49,760 and why do all of these things need to exist? 13 00:00:49,760 --> 00:00:53,260 Why wouldn't one programming language, for example, be enough? 14 00:00:53,260 --> 00:00:55,400 A related question, but slightly different is, 15 00:00:55,400 --> 00:00:57,360 why are there new programming languages? 16 00:00:57,360 --> 00:01:00,820 Given that we have so many programming languages already, 17 00:01:00,820 --> 00:01:04,200 what is the need for new ones to be created? 18 00:01:05,840 --> 00:01:09,040 Finally, how do we know a good programming language when we see it? 19 00:01:09,040 --> 00:01:12,671 What makes a good programming language, and what makes a bad programming language? 20 00:01:12,671 --> 00:01:18,964 I just wanted to spend, this video anyway, talking about these three questions. 21 00:01:18,964 --> 00:01:24,038 As we'll see, I think the answers to these questions are largely 22 00:01:24,038 --> 00:01:28,744 independent of the technical aspects of language design and 23 00:01:28,744 --> 00:01:33,463 implementation, but very interesting in their own right. 24 00:01:33,463 --> 00:01:38,259 Let's begin with the question of why are there some many programming languages, and 25 00:01:38,259 --> 00:01:42,066 at least a partial answer to this question is not too hard to come by. 26 00:01:42,066 --> 00:01:46,087 If you think for a few minutes, you'd realize that the application domains for 27 00:01:46,087 --> 00:01:49,720 programming have very distinctive and conflicting needs. 28 00:01:49,720 --> 00:01:55,510 That is, it's very hard to design one language that would actually do everything 29 00:01:55,510 --> 00:02:00,820 in every situation for all programmers, and let's just go through some examples. 30 00:02:00,820 --> 00:02:06,926 One domain that you might not think about very much is scientific 31 00:02:06,926 --> 00:02:12,170 computing, so these are all the big calculations that are done for 32 00:02:12,170 --> 00:02:16,716 engineering applications primarily, but also big science and 33 00:02:16,716 --> 00:02:23,070 long-running experiments, simulation experiments. 34 00:02:23,070 --> 00:02:26,360 What are the needs for such computations? 35 00:02:26,360 --> 00:02:30,250 Well, typically, you need very good floating point support. 36 00:02:30,250 --> 00:02:31,290 I'll abbreviate that as FP. 37 00:02:32,360 --> 00:02:35,720 You need good support for arrays and 38 00:02:35,720 --> 00:02:40,873 operations on arrays because the most common data type 39 00:02:40,873 --> 00:02:45,691 in most scientific applications is large arrays of 40 00:02:45,691 --> 00:02:51,196 floating point numbers, and you also need parallelism. 41 00:02:51,196 --> 00:02:54,345 Today to get sufficient performance you really have to 42 00:02:54,345 --> 00:02:57,220 exploit parallelism in these applications, and 43 00:02:57,220 --> 00:03:00,926 not every language actually supports all of these things well. 44 00:03:00,926 --> 00:03:03,839 This is actually not an exhaustive list of the things you need, but 45 00:03:03,839 --> 00:03:06,600 it's a few distinctive things that are needed. 46 00:03:06,600 --> 00:03:11,240 One language that has traditionally done a very good job of supporting these things 47 00:03:11,240 --> 00:03:16,090 is Fortran, and Fortran is still heavily used in the scientific community. 48 00:03:16,090 --> 00:03:18,940 It was originally designed for scientific applications. 49 00:03:18,940 --> 00:03:26,090 If you recall, the name means formula translation, and it has evolved over time. 50 00:03:26,090 --> 00:03:29,090 It doesn't really look much like the original language anymore, 51 00:03:29,090 --> 00:03:32,950 but it's always retained this core constituency in scientific computing and 52 00:03:32,950 --> 00:03:36,040 remains one of the leading languages in that domain. 53 00:03:37,250 --> 00:03:43,044 Now, a completely different kind of domain is business applications. 54 00:03:47,107 --> 00:03:48,270 What do you need here? 55 00:03:48,270 --> 00:03:51,050 Well, so here you're going to need things like persistence. 56 00:03:52,620 --> 00:03:55,010 You don't want to lose your data. 57 00:03:55,010 --> 00:03:57,440 Businesses go to a lot of trouble to get the data, and 58 00:03:57,440 --> 00:04:03,470 they need a way to hold onto it, and they want that to be extremely reliable. 59 00:04:03,470 --> 00:04:08,150 You're going to need good report facilities because, typically, 60 00:04:08,150 --> 00:04:12,490 you want to do something with the data, so you need good facilities for 61 00:04:12,490 --> 00:04:17,670 report generation, and, also, you want to be able to exploit the data. 62 00:04:17,670 --> 00:04:20,970 The data is actually, in many modern businesses, one of the most valuable 63 00:04:20,970 --> 00:04:26,630 assets, and so you need good facilities for asking questions about your data. 64 00:04:26,630 --> 00:04:28,900 Let's call it data analysis. 65 00:04:28,900 --> 00:04:32,710 Again, this is not an exhaustive list of things that you need, 66 00:04:32,710 --> 00:04:35,260 but it is representative, I would say. 67 00:04:36,520 --> 00:04:41,645 Probably the most common, or one of the most commonly used languages for 68 00:04:41,645 --> 00:04:46,815 this class of applications, is SQL, the database query language. 69 00:04:46,815 --> 00:04:52,255 Relational databases and their associated programming languages, 70 00:04:52,255 --> 00:04:57,975 but most notably SQL, really dominate in this application domain. 71 00:04:59,155 --> 00:05:03,347 Then, another domain, let's do one more, is systems programming. 72 00:05:06,423 --> 00:05:10,845 By this, I mean things like embedded systems, things that control devices, 73 00:05:10,845 --> 00:05:15,410 operating systems, things like that, and what are the characteristics here? 74 00:05:15,410 --> 00:05:19,440 Well, we need very low level control of the resources. 75 00:05:19,440 --> 00:05:24,220 The whole point of systems programming is to do a good job of managing resources, 76 00:05:24,220 --> 00:05:28,510 and so we really want fine grain control over the resources. 77 00:05:30,340 --> 00:05:34,664 Often there's a time aspect, so you might have some real-time constraints, so 78 00:05:34,664 --> 00:05:36,805 you need to be able to reason about time. 79 00:05:40,544 --> 00:05:43,595 Because these are actually, again, devices, and 80 00:05:43,595 --> 00:05:47,560 they need to be able to react within certain amounts of time. 81 00:05:47,560 --> 00:05:50,010 If it's a network device or something like that, 82 00:05:50,010 --> 00:05:53,590 you need to be responsive to the network, lots and lots of things. 83 00:05:53,590 --> 00:05:58,120 Lots and lots of examples where timing is important, and these are just two aspects, 84 00:05:58,120 --> 00:06:00,750 and I'm running out of space here, so I'll just stop with that. 85 00:06:00,750 --> 00:06:04,080 But again, these are representative of the kinds of things you need 86 00:06:04,080 --> 00:06:05,110 in systems programming. 87 00:06:06,260 --> 00:06:10,200 Probably today, still the most widely used systems programming language, or 88 00:06:10,200 --> 00:06:14,300 family of systems programming languages, is the C and, 89 00:06:14,300 --> 00:06:17,672 to some extent, C++ family of languages. 90 00:06:19,330 --> 00:06:22,920 As you can see, the requirements in these different domains are just completely 91 00:06:22,920 --> 00:06:24,203 different from each other. 92 00:06:24,203 --> 00:06:28,105 What's important in one domain, or most important in one domain, 93 00:06:28,105 --> 00:06:31,732 is not the same as in another domain, and it's easy, I think, 94 00:06:31,732 --> 00:06:36,044 to imagine at least that it would be difficult to integrate all of these into 95 00:06:36,044 --> 00:06:39,351 one system that would do a good job on all of these things. 96 00:06:43,575 --> 00:06:45,310 That brings us to our second question. 97 00:06:45,310 --> 00:06:47,540 Why are there new programming languages? 98 00:06:47,540 --> 00:06:50,000 Okay. There are so many languages in existence. 99 00:06:50,000 --> 00:06:52,760 Why would we ever need to design a new one? 100 00:06:54,190 --> 00:06:57,380 I'm going to begin the answer to this question with an observation that, 101 00:06:57,380 --> 00:07:01,670 at first glance, has nothing to do with the question at all, so 102 00:07:01,670 --> 00:07:03,220 let me just take a moment to explain it. 103 00:07:03,220 --> 00:07:07,360 I claim that programmer training is the dominant cost for a programming language, 104 00:07:07,360 --> 00:07:08,710 and I think this is really important, 105 00:07:08,710 --> 00:07:13,500 so I'm just going to emphasize the bit that's important here. 106 00:07:13,500 --> 00:07:14,940 It's the programmer training. 107 00:07:14,940 --> 00:07:18,300 The cost of educating the programmers in the language. 108 00:07:18,300 --> 00:07:20,580 If you think about a programming language, 109 00:07:20,580 --> 00:07:24,660 there are several things that have to happen for that language to get used. 110 00:07:24,660 --> 00:07:27,570 Somebody has to design it, but that's really not very expensive. 111 00:07:27,570 --> 00:07:31,030 That's just one or a very few people, typically. 112 00:07:31,030 --> 00:07:35,640 Somebody has to build a compiler, but that is also not actually all that expensive. 113 00:07:35,640 --> 00:07:37,310 Maybe 10 to 20 people for 114 00:07:37,310 --> 00:07:40,740 a really large compiler project can build quite a good compiler. 115 00:07:41,890 --> 00:07:45,387 The real cost is in all the users, in educating them. 116 00:07:45,387 --> 00:07:48,692 If you have thousands or hundreds of thousands or 117 00:07:48,692 --> 00:07:51,754 millions of users of a language, the time and 118 00:07:51,754 --> 00:07:57,095 money that it takes to teach them all the language is really the dominant cost. 119 00:07:57,095 --> 00:08:01,375 Here I don't mean just the actual dollar expense of buying textbooks and 120 00:08:01,375 --> 00:08:04,140 taking classes and things like that. 121 00:08:04,140 --> 00:08:07,900 It's also the fact that the programmers have to decide that it's worth it for 122 00:08:07,900 --> 00:08:12,640 them to learn this language, and many programmers learn on their own time, but 123 00:08:12,640 --> 00:08:14,095 that's a use of their time. 124 00:08:14,095 --> 00:08:19,180 The expense of their time is a real economic cost, and so, if you think about 125 00:08:19,180 --> 00:08:23,680 the number of hours that it takes to teach a population of a million programmers 126 00:08:23,680 --> 00:08:29,019 a language, that's really quite a significant economic investment. 127 00:08:30,040 --> 00:08:30,700 All right. 128 00:08:30,700 --> 00:08:35,080 Now, from this observation, we can make a couple of predictions pretty easily. 129 00:08:35,080 --> 00:08:37,980 Again, these are just predictions, 130 00:08:37,980 --> 00:08:43,080 now, that follow from this claim if you believe that it's true. 131 00:08:43,080 --> 00:08:48,463 Let me erase that and fix it, so first prediction 132 00:08:48,463 --> 00:08:54,651 is that widely-used languages will be slow to change. 133 00:09:01,373 --> 00:09:02,650 Why should that be true? 134 00:09:02,650 --> 00:09:07,100 Well, if I make a change to a language that lots of people use, 135 00:09:07,100 --> 00:09:11,730 I have to educate everybody in that community about the change. 136 00:09:11,730 --> 00:09:15,310 Even relatively minor language extensions or 137 00:09:15,310 --> 00:09:19,740 small changes to syntax small new features, even just simple changes in 138 00:09:19,740 --> 00:09:22,560 the interface of the compiler, if you have a lot of users, 139 00:09:22,560 --> 00:09:28,250 it takes a very long time and it's quite expensive to teach them all about that. 140 00:09:28,250 --> 00:09:35,420 As languages become widely used, their rate of change will slow down. 141 00:09:36,670 --> 00:09:40,700 This predicts that over time, as the world of programming grows, as we have more and 142 00:09:40,700 --> 00:09:43,960 more programmers in the world, we would expect the most popular languages, 143 00:09:43,960 --> 00:09:48,780 which will have larger and larger user bases, larger and larger programmer bases, 144 00:09:48,780 --> 00:09:53,620 to become more and more ossified, to evolve more and more slowly. 145 00:09:53,620 --> 00:09:54,340 I think, actually, 146 00:09:54,340 --> 00:09:57,830 what you see in practice is very consistent with that prediction. 147 00:09:59,080 --> 00:10:04,120 Now, at the other end of the spectrum, this same observation 148 00:10:04,120 --> 00:10:08,640 makes an almost what appears to be a contradictory prediction, 149 00:10:08,640 --> 00:10:12,670 which is that it's easy to start a new language. 150 00:10:16,600 --> 00:10:21,730 That, in fact, the cost of starting up a new language is very low, and why is that? 151 00:10:21,730 --> 00:10:26,420 Well, because you start zero users, and so there's essentially zero training cost 152 00:10:26,420 --> 00:10:29,730 at the beginning, and then, even when you have just a few users, 153 00:10:29,730 --> 00:10:35,490 the cost of teaching them the changes in the language is not very high. 154 00:10:35,490 --> 00:10:38,960 New languages can evolve much more quickly. 155 00:10:38,960 --> 00:10:43,342 They can adapt much more quickly to changing situations. 156 00:10:43,342 --> 00:10:48,560 It's just not very costly to experiment 157 00:10:48,560 --> 00:10:52,485 with a new language at all, and there's a tension between these two things. 158 00:10:52,485 --> 00:10:55,415 Okay? 159 00:10:55,415 --> 00:11:00,605 When is a programmer going to choose between a widely used existing language 160 00:11:00,605 --> 00:11:05,319 that perhaps doesn't change very quickly and a brand new language? 161 00:11:05,319 --> 00:11:09,358 They're going to choose it if their 162 00:11:09,358 --> 00:11:14,620 productivity now exceeds the training cost. 163 00:11:16,500 --> 00:11:19,610 If they perceive that by spending a little bit of time and 164 00:11:19,610 --> 00:11:23,848 money to learn this new language they're going to be much more productive over 165 00:11:23,848 --> 00:11:27,970 a relatively short period of time, then they're going to make the switch. 166 00:11:27,970 --> 00:11:29,107 Okay? 167 00:11:29,107 --> 00:11:31,400 When is this likely to happen? 168 00:11:32,610 --> 00:11:39,006 Well, putting this all together, languages are most likely to be adopted. 169 00:11:44,148 --> 00:11:45,379 To fill a void. 170 00:11:50,080 --> 00:11:54,220 Okay, and, again, this is a prediction that follows from the fact of programmer 171 00:11:54,220 --> 00:11:55,606 training is the main cost. 172 00:11:55,606 --> 00:11:57,326 What do I mean by this? 173 00:11:57,326 --> 00:12:03,105 Well, what I mean is that programming languages exist for a purpose. 174 00:12:03,105 --> 00:12:05,165 People use them to get work done, and 175 00:12:05,165 --> 00:12:08,953 because we're still in the middle of the information revolution, 176 00:12:08,953 --> 00:12:12,417 there are new application domains coming along all the time. 177 00:12:12,417 --> 00:12:17,381 There are new kinds of programming that emerge every few years, or 178 00:12:17,381 --> 00:12:19,970 even more often than that. 179 00:12:19,970 --> 00:12:21,940 In terms of recent history, 180 00:12:21,940 --> 00:12:25,660 mobile applications are now something that's relatively new, and 181 00:12:25,660 --> 00:12:30,030 there's a lot of new technology built up to support mobile computing. 182 00:12:30,030 --> 00:12:34,010 A few years ago it was the Internet itself was the new programming platform, and 183 00:12:34,010 --> 00:12:36,935 a bunch of new programming languages like Java, in particular, 184 00:12:36,935 --> 00:12:41,750 got started during that time. 185 00:12:41,750 --> 00:12:46,460 New programming niches open up because the technology changes, what people want to 186 00:12:46,460 --> 00:12:51,360 do with software changes, and this creates new opportunities for languages. 187 00:12:51,360 --> 00:12:56,410 The old languages are slow to change, and so they have some difficulty 188 00:12:58,050 --> 00:13:02,040 adapting to fit these new domains. 189 00:13:02,040 --> 00:13:05,610 They aren't really necessarily well-suited to them for the reasons we talked about 190 00:13:05,610 --> 00:13:10,480 on the previous slide with the previous question because it's hard to have 191 00:13:10,480 --> 00:13:16,180 one language that incorporates all the features you would want. 192 00:13:16,180 --> 00:13:19,920 The new languages are not necessarily perfect for these application domains. 193 00:13:19,920 --> 00:13:24,360 They're slow to adapt to the new situation, and 194 00:13:24,360 --> 00:13:26,490 this tends to call forth new languages. 195 00:13:26,490 --> 00:13:31,410 When there's a new opportunity in some application domain, if there 196 00:13:31,410 --> 00:13:35,329 are enough programmers to support the language, often a new language will arise. 197 00:13:37,570 --> 00:13:41,560 I just want to point out another prediction that can be made from this 198 00:13:41,560 --> 00:13:46,251 one observation, that programmer training, again, I'll underline that, 199 00:13:46,251 --> 00:13:49,206 is the dominant cost for a programming language. 200 00:13:49,206 --> 00:13:51,189 That is that new languages. 201 00:13:55,684 --> 00:13:57,343 Tend to look like old languages. 202 00:14:06,665 --> 00:14:10,700 That is, that new languages are rarely, if ever, completely new. 203 00:14:10,700 --> 00:14:15,940 They have a family resemblance to some predecessor language, 204 00:14:15,940 --> 00:14:19,890 sometimes a number of predecessor languages, and why is that? 205 00:14:20,950 --> 00:14:24,490 Well, partly that it's hard to think of truly new things, but also, I think that 206 00:14:24,490 --> 00:14:29,280 there's an economic benefit to this, namely that it reduces the training cost. 207 00:14:29,280 --> 00:14:31,760 By having your new language look like an old language, 208 00:14:31,760 --> 00:14:35,380 by leveraging off what people already know about the old language, 209 00:14:35,380 --> 00:14:37,520 you make it easier for people to learn the new language. 210 00:14:37,520 --> 00:14:39,640 You make them learn it more quickly, and 211 00:14:39,640 --> 00:14:44,990 the most classic example of this is Java versus C++. 212 00:14:44,990 --> 00:14:49,670 Where Java was designed to look a lot like C++, and 213 00:14:49,670 --> 00:14:53,520 that was, I think, very conscious to make it easy for 214 00:14:53,520 --> 00:14:57,210 all of the existing C++ programmers to start programming in Java. 215 00:15:00,960 --> 00:15:04,922 Finally, we can ask ourselves, what is a good programming language? 216 00:15:04,922 --> 00:15:09,623 Here, unfortunately, the situation is much less clear. 217 00:15:09,623 --> 00:15:16,652 I would make just one claim, that there is no, and I'll emphasize no, 218 00:15:16,652 --> 00:15:21,820 universally accepted metric for language design. 219 00:15:26,590 --> 00:15:28,020 What do I mean by that? 220 00:15:28,020 --> 00:15:33,820 Well, I guess, the most important part of this 221 00:15:33,820 --> 00:15:37,980 statement is the universally accepted bit, so 222 00:15:37,980 --> 00:15:41,630 I mean that people don't agree on what makes a good language. 223 00:15:41,630 --> 00:15:45,580 There are lots of metrics out there, and people have proposed lots of ways of 224 00:15:45,580 --> 00:15:50,800 measuring programming languages, but most people don't 225 00:15:50,800 --> 00:15:55,750 believe that these are very good measures, and there is certainly no consensus. 226 00:15:55,750 --> 00:15:58,430 If you just look at the world of programmers, 227 00:15:58,430 --> 00:16:02,050 they can't agree on what the best language is, and to convince yourselves of this, 228 00:16:02,050 --> 00:16:07,007 just go and take a look at any of the many Newsgroup posts where people get into 229 00:16:07,007 --> 00:16:11,220 semi-religious arguments about why one group of languages, or 230 00:16:11,220 --> 00:16:14,159 particular language, is better than another language. 231 00:16:15,670 --> 00:16:19,755 Even in the research community, in the scientific community and 232 00:16:19,755 --> 00:16:22,949 among the people who design languages, I would say 233 00:16:22,949 --> 00:16:27,937 that there is no universally accepted consensus on what makes a good language. 234 00:16:27,937 --> 00:16:31,699 To just kind of illustrate the difficulties in trying to come up with 235 00:16:31,699 --> 00:16:35,461 such a metric, let me discuss one that I've heard people propose, 236 00:16:35,461 --> 00:16:38,502 in all seriousness, and that is that a good language. 237 00:16:42,971 --> 00:16:44,570 Is one people use. 238 00:16:50,400 --> 00:16:54,400 Let me put a question mark on that because I don't believe this statement, 239 00:16:56,320 --> 00:17:00,130 and I think, with a moment's reflection, 240 00:17:00,130 --> 00:17:04,140 I can convince you that this isn't a great measure. 241 00:17:04,140 --> 00:17:06,770 On the positive side, I guess, the argument for 242 00:17:06,770 --> 00:17:10,060 this is that it's a very clear measure. 243 00:17:10,060 --> 00:17:11,770 It measures the popularity of the language. 244 00:17:11,770 --> 00:17:14,319 How many people are actually using it, and presumably 245 00:17:15,450 --> 00:17:19,380 languages that are more widely used are more widely used for a good reason. 246 00:17:19,380 --> 00:17:24,323 In some sense, perhaps, they are better languages, but this would imply, 247 00:17:24,323 --> 00:17:28,203 if you believe this and follow it to its logical conclusion, 248 00:17:28,203 --> 00:17:30,805 that Visual Basic is the best language. 249 00:17:34,486 --> 00:17:38,027 Yeah, above all other programming languages, and 250 00:17:38,027 --> 00:17:41,020 I have nothing against Visual Basic. 251 00:17:41,020 --> 00:17:45,700 It's a well-designed system, but I don't even think the designers of Visual Basic 252 00:17:45,700 --> 00:17:50,820 would claim that it is, in fact, the world's best programming language. 253 00:17:50,820 --> 00:17:55,110 As we saw in the discussion that we just had, there are many, 254 00:17:55,110 --> 00:17:58,860 many other factors besides technical excellence 255 00:17:58,860 --> 00:18:03,230 that go into whether a programming language is widely used or not. 256 00:18:03,230 --> 00:18:07,820 In fact, technical excellence is probably not even the most important reason 257 00:18:07,820 --> 00:18:09,330 that a language might be used. 258 00:18:09,330 --> 00:18:15,030 It has much more to do with whether it addresses a niche or 259 00:18:15,030 --> 00:18:18,410 an application domain for which there isn't a better tool. 260 00:18:18,410 --> 00:18:21,430 Then, once it's established and has lots of users, of course, there's inertia and 261 00:18:21,430 --> 00:18:23,960 history that aid it in surviving. 262 00:18:23,960 --> 00:18:30,560 That's why we still have FORTRAN and COBOL and lots of other languages from long, 263 00:18:30,560 --> 00:18:34,600 long ago that we could, if we were starting over today, design much better. 264 00:18:37,020 --> 00:18:40,280 To conclude this video on the economy of programming languages, 265 00:18:40,280 --> 00:18:44,360 I think the two most important things to remember are that application domains have 266 00:18:44,360 --> 00:18:47,770 conflicting needs, and, therefore, 267 00:18:47,770 --> 00:18:52,440 it's difficult to design one system 268 00:18:54,640 --> 00:18:57,590 that incorporates everything that you would like to have. 269 00:18:59,470 --> 00:19:04,030 You can't get all the features that you would like into 270 00:19:04,030 --> 00:19:08,070 a single system in a coherent design, at least it's very hard to do that, and so 271 00:19:08,070 --> 00:19:14,320 it takes a lot of time to add new features to existing systems. 272 00:19:14,320 --> 00:19:18,844 The second point is that programmer training is the dominant cost for 273 00:19:18,844 --> 00:19:24,382 a programming language, and together these two things, these two observations, 274 00:19:24,382 --> 00:19:28,360 these really explain why we get new programming languages. 275 00:19:28,360 --> 00:19:33,026 Because the old languages are difficult to change, and when we have new 276 00:19:33,026 --> 00:19:38,472 opportunities, it's often easier and more direct to just design the language for 277 00:19:38,472 --> 00:19:43,218 those rather than trying to move the entire community of programmers and 278 00:19:43,218 --> 00:19:47,143 existing systems to accommodate those new applications.