Oral History of linus Torvalds
Yüklə 332,77 Kb. Pdf görüntüsü
|
Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 36 of 41
Torvalds: Yes. And it's actually-- well, I'm driving it in the sense that I'm the one who says, this is when we do the first release candidate. And then I decide usually two months afterwards, okay, this is released, but it's all based on time. We don't…
Torvalds: Right now we're on 2.6.26 and we're working towards 2.6.27. And I'm actually in the end part of the current to merge window so. Booch: How often are you scheduling releases? What feels right? Torvalds: It's not what feels right, it's what works. And what we found is what works is two weeks of merge window because one week is too short of a time. Because that means that everybody has to synchronize and everybody has to have everything lined up just right and nobody can really conflict. So, one week isn't enough to merge, but more than two weeks would be too much because then people have time to start trying to do a new development and they're not supposed to do that. They're supposed to just merge what they have already worked on. And then two months of stabilization because longer than two months people start losing interest. And shorter doesn't work, so.
update the major number just to make the minor numbers be smaller, right? That's literally how it's working. We have stopped doing releases based on features. We do releases based on time and the features that come in are the features that are ready. It just doesn't work to have a plan saying, okay, the next release needs to have these five things because that is what people are asking for. It doesn't work because it turns out that out of the five things, only three are ready. One turned out to be a really bad idea to begin with, and one might make it to the next release, but probably won't, kind of thing. So, if you do a release based on features, you're stuck and you end up with a mess that Microsoft is in because they apparently do that. And then you have release cycles that are three, four years long. We used to do that and we stopped just because it doesn't work. Booch: What's a typical release cycle from beginning to end? Torvalds: The release cycle is literally tw o weeks of merging and then…
basically. I aim for two months. Two-and-a-half months is what I consider to be realistic and then bad releases when we can't fix it up in time, take over three months. So, that's why we don't actually get four releases a year. Booch: So just a handful of a few more topics. What is Linux going to look like five years, ten years from now? Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 37 of 41
Torvalds: I don't even think that way. Booch: Really? Torvalds: No, no because most of the things that push us are external. There are very few things that, okay, we have this thing that we want to push on people. It works the other way around where the whole point of an OS, as far as I'm concerned, is really to be the layer between the hardware and the real software. The OS doesn't do anything on its own. That's my personal opinion. That's how it should be. And if your OS does something on its own, you're doing something wrong. And that means that the OS depends mostly on how hardware develops, which clearly happens all the time, right? So, that's not going to end any time soon. But also very much how people use it and what capabilities the programs need from the hardware that the OS just has to kind of juggle. Sometimes it's not hardware, sometimes it's just prediction domains between two pieces of software. But that just means that we don't look for features we want to implement. We look for features we can do better that we already have or more commonly, we just have input from hardware or from users saying, hey, we have so and so problem and I can't do this. And that's what tends to drive us. Booch: Are multi-core chips sort of pushing you in a certain way right now? Torvalds: Well, right now I don't think it's realistic to-- I mean, it's-- we pushed our scalability up to-- I mean, people have been selling Linux machines with 512 CPUs for a long time and in some cases more. We're pushing that up, but it's-- so we have done patches to merge it so that it works better on… if we 4,000 CPUs and we'll continue to push that up to 8,000 CPUs. But realistically, that's still going to be the special case. Booch: Multi-core is a special case? Torvalds: Well, not multi-core per say, but going into the thousands. Most people would have four or eight and that's kind of-- we've shifted what we optimize for over time. It used to be we optimized for a single CPU core. And, okay, so that's not realistic even in the embedded world we have-- two is pretty much the norm now. In a couple of years four will be the norm. My new machine has eight. So, it's not a huge issue and I don't think it's realistic to-- software-- doing threaded software is too hard.
how the brain works, but it's not actually how you think as a programmer. Very few programmers can really think in a parallel world. It's really hard to do.
languages? Torvalds: I think it's a limitation of the human brain because it's a limitation of our language in the sense of our human language. We are linear thinkers because we think in languages. We don't think in some parallel model, right? At least our conscious thinking is linearized by our linear language view, which means that fundamentally the only thing where you get parallelism is if the problem is really parallel or if Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 38 of 41
you find a way to turn human linear thinking into parallel thinking. And people have been trying to do that for the last 30 years and have generally been very unsuccessful at it. So, it's true that certain problems and, I mean, video and graphics in general tend to be extremely parallel and so trivially at parallel that it's easy to think about the parallelism. But I'm not a huge believer in the multi-core as a general purpose of view kind of thing. So, we'll see what happens. I've been wrong before, but I don't think I am when it comes to this. Booch: What are two of the things you've been wrong about? Torvalds: Okay, the really embarrassing one and which is kind of related to this one was I did not think SMP was going to be realistic on the desktop at all ten years ago. And the reason was is that it was also expensive. And then Intel turned the tables and just integrated everything on their die and suddenly doing two CPUs was almost as cheap as doing just one. So, hardware advances do change and I-- who knows, I may be wrong about people being able to take advantage of many, many cores.
hand, there haven't been any things I've been really unhappy about either because we can fix the technical side. So, maybe the thing I've been really, really right about was really the whole license choice. And so I think that has been the thing that really mattered most. Booch: What makes Linus happy? Where are your passions? Torvalds: I don't know. I've been really happy about the fact that my job is my hobby, right, or my hobby is my job. However you want to put it, right. And the fact that what I do is not just fun to do, but it also seems to be meaningful, so it actually matters to other people. And if you have those two things, I think you'll be happy whatever you do. I mean, it doesn't have to be computers, but within computers I've-- that's what I ask for. That's what I hope my kids will have, although I don't know what they…
Torvalds: I don't know. It doesn't really seem to be that way, but I don't care. It's not like I'm going to push them that way either. I just hope that they find something that they are really interested in. Booch: In fact, what advice would you give to the high school or college student who wants to follow their passion? Do you have any words of wisdom for them? Torvalds: Well, if you want to follow your passion, do it. Right? But on the other hand, I mean, a lot of-- that's a dangerous thing to say because a lot of people's passions are really stupid. I mean, let's face it. It works when your passion makes sense, but quite often your passion isn't really going to be your work anyway and there's no way. I mean, you may be passionate about any number of things that just aren't worthwhile.
Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 39 of 41
Torvalds: Yeah. I mean, so, it's kind of-- it worked for me, but I'm not going to say it's going to work for anybody else. Booch: So, you actually may have already answered this, but if you hadn't sort of discovered software, what would you have been doing do you imagine or can you not imagine a life without this? Torvalds: I don't know. I have been doing programming for so long that it's kind of hard to imagine what else I would do. I mean, I know what I'm interested in and I would probably-- if it hadn't been computer science, my guess would be I'd be doing biology. And especially something like I'm to do with DNA and evolution, and like biology on a low level, the same way I enjoy doing computers at a low level. I would not be the kind of person who actually deals with the animals. I would be the kind of person who deals with cells and then things like that, I think. Booch: You have to be proud to realize Linux probably runs most of the DNA sequencing machines these days. Torvalds: Yeah. Well, except the ones that are special built and don't have any OS at all because they're just doing the calculations, yeah. But yeah that's…
and biology and the brain. Torvalds: When I was younger I used to read a lot about physics and…
was apparently not a very nice person, he was certainly one of those people who changed the world in many different ways. And, I mean, Einstein obviously, so that was physics. And I mean, I wasn't that aware of Darwin as a big mind because in Finland I was never-- I mean, that was never in doubt, no. This whole American school system is kind of broken.
— but I ended up being slightly-- I noticed when I actually did physics at a university that I wasn't as interested in it as I thought I had been. And it wasn't-- it just wasn't as interesting. It probably needed a bigger effort than I put into. So, besides being more about psychology and brain development and evolution and things like that, that I actually feel-- if I read some non-fiction books these days, it tends to be about human behavior or evolution. Booch: Like what's the last non-fiction book that comes to mind that you've read or the last fiction book for that matter that you've read? Torvalds: Well, I guess the last non-fiction was actually Michael Pollan's In Defense of Food, actually. So, that didn't actually fit that particular pattern at all. But it might have been Oxygen by-- I forget who the Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 40 of 41
author. But it's about, well, at least some of the theories about how the presence of oxygen actually impacted evolution on earth. Booch: So, next to the last question is out of total curiosity. Not counting spam, how much email do you process on a daily basis? Torvalds: It's not that bad actually. So, A, I don't get any spam because my spam filters are extremely aggressive so. I mean, I get a few pieces of spam, but it's really not even noticeable. And I'm on a few lists, but they all go to a separate mailbox. So, my mailing lists end up being things that I look up if I need to, but I don't normally read them at all. So, the only email I get are literally emails that are sent to me personally, and I get too much of that. But I found that not answering email that is unnecessary is a great way to kind of push back the flow of it. Because it turns out that the people who really need to contact me either can just repeat the email or their question in the email was interesting and relevant enough that I actually answered it in the first place. It does mean that I actually, like, read a lot more email than I answer because my answering email is kind of-- it's a failure when I answer emails. Well, not always, I mean. Quite often it's because I'm the person who needs to answer that email, but it's something I've actually tried actively not to do.
email is so much better. And if you have to have a phone, you should just send SMS messages. But that doesn't work so well in the U.S. because it's too expensive and not everybody can take them anyway, right, there either. So, I think phones are a horrible distraction. And part of it is the synchronous nature of it, so you had to kind of-- I'm in the middle of something and the phone rings, and I hate it. The whole switching between two tasks. I'm good at concentrating, but the downside of that is when something disturbs my concentration, I really hate it and phones do that. Cell phones are just phones that are designed to be even more annoying. That's probably one of the reasons I had to move away from Finland, because if I hadn't moved away they'd probably deport me because Nokia is-- the whole Finnish economy revolves around cell phones almost. Booch: And now the Lima [ph?] Foundation put Linux on the cell phones. Torvalds: Yeah, well, yeah. I feel a bit bad about that. Booch: Why do you feel bad about that? Torvalds: No, I'm kidding. I wouldn't use…
Oral History of Linus Torvalds
CHM Ref: X4147.2008 © 2008 Computer History Museum Page 41 of 41
Torvalds: I can avoid the cell phones. If other people want to use cell phones, say, it's all good as long as I don't have to. Booch: So, the last question is are there other questions I should have asked you? Torvalds: No. No, I don't know. Booch: Well, thank you very much for your time. You are a fascinating individual. And most importantly you're having a hell of a lot of fun and I envy that. Torvalds: Yeah, it's a good life.
END OF INTERVIEW Yüklə 332,77 Kb. Dostları ilə paylaş:
|