Conversation Linux kernel daemon AndrewMorton

Dialog with the Linux kernel daemon AndrewMorton-Linux general technology-Linux programming and kernel information. The following is a detailed description. HOST: Thank you, Jonathan Corbet. Next we will ask Andrew Morton, the guardian of Linux kernel 2.6, to have a conversation.

Jonathan Corbet: I have a few questions that can be discussed with you. If you have any questions, you can also answer your questions. First of all, there is a question that I mentioned just now, we have a relatively high level of Pach in the kernel. I would like to ask Andrew Morton how you can solve this problem, that is, how you solve a relatively high level of Pach, at the same time, it maintains its kernel.

Andrew Morton: Actually, I am not managing it. It doesn't mean that you manage as much as you think. In fact, people sometimes depend on me. In addition, I sometimes do not conduct many such reviews on him. If this proportion is relatively high, I will tell others that I am sorry I don't have the time to deal with this, at the same time, I often reject requests from other R & D personnel because I really don't have that much time.

Jonathan Corbet: you attended a meeting two years ago. You said it would slow down because you thought it was almost done. So in the past few years, I think it was indeed like this, so what do you think you want? Do you want it to slow down, and there is such a high-speed change? Why? Why is it constantly changing at such a high speed?

Andrew Morton: I think the number of involved companies is related to applications. We now have a lot of R & D personnel who have more requirements on the kernel, I think this is why we have such a high replacement rate or a high update rate, and new technologies are constantly emerging, as well as new hardware technologies, all require more support, but I think the last kernel update speed will slow down, because from the most fundamental change, for example, a new CPU, new R & D and so on. However, some kernel technologies will be more mature in the future. So in the future, this will stabilize and slow down.

Jonathan Corbet: What changes do you think you need before it slows down. What do you think we need more work now.

Andrew Morton: people often ask me this question. I don't know what new content the kernel will propose in the next six months, because we don't have a unified control mechanism in the kernel, we don't have a roadmap, we don't have a plan, so I don't know. I can only talk about every company, everyone will contribute to Linux. So people often tell us what they are doing, so I am just guessing, but in the long term, I don't know what people think in their minds, I have no way to say which direction it will take for a long time. I think the best prediction is that I generally read your report. For example, I know a lot from your speech.

Jonathan Corbet, so it is possible that the results will be visible only after a new version is released. I think the best change you have brought about is that we have established a very positive mechanism, that is to say, people will hear a feedback after putting his things up, so the Pach they put up will not be forgotten, etc, I would like to ask what kind of reflection is like for Pach. If I contribute something to the Internet and no one gives it to me, then what should I do?

Andrew Morton: I always want to ensure that some contributions of everyone I get will respond. Especially for some of their achievements, if you send them, if others have comments but I have not commented on them, that is to say, I agree with others. If I disagree with others, I will give you a response or feedback. For example, I had a two-week holiday for Christmas, but I still had 500 Pach. So I kept working on it. There are also some possibilities. If you think you have put it there and you don't think anyone cares about you, then you will send me a private email and I will give you some feedback. In general, I will tell you, for example, whether some people disagree or not, and I will give a feedback.

Jonathan Corbet: When people are thinking about joining our community, it is very difficult for them to find a way to interact with our community. So I have a question, how Should new developers join the community to become part of the community? How can we avoid errors?

Andrew Morton: I think one thing people must be careful about is that we use e-mail to do our work. If you want someone to always respond to you, if it is a very appropriate response, it is very difficult for a person who has not responded for a week. Therefore, it is very important to receive emails from others. We usually have this problem. They will ask me how to start this kernel development. I think the best answer is that you don't want to ask this question, or you just need to ask this question to start development. First, you need a good English reading ability. You need to write English, network connection and computer, and then start kernel development and connect your computer to the network, download the kernel and start. Join the community and get familiar with the community. The next step is to download the latest kernel version. If you download my version, you may find some defects and errors. It is a very simple process to communicate with other developers and solve these problems. After two or three months, you may have become a kernel developer and a member of this community, so the most important thing is to start. Once we start working for two or three months, we will be very familiar with the entire environment, next, you need to consider what kind of original and original work you need to contribute for yourself and send some changes, then you have your own name in the entire change database, so people can find your name and see what you have done. This is a very correct way to start your own kernel development career.
In some companies, the practice is to expand a group of developers. In this way, there are actually many competitors who need people to interact when people participate in the entire process, many people are very polite and conservative when launching Pach. What kind of etiquette do we deal with each other? I think if I receive a very rude reply, then going back is to speak on behalf of others, because each person does not necessarily speak on his own. One person said a sentence, or it has caused opposition from thousands of people. This is a way of working. If a person says a word so everyone agrees, this may lead to a successful job, but in more cases it may be opposed by thousands of people, do not do this. In fact, there are some very famous people who do not speak on behalf of everyone. They only speak on behalf of themselves. In fact, this is also a problem.

Jonathan Corbet: I hope I can agree with you. Another problem that developers encounter is that some people are very fast at the time of making changes, and some people can finish the changes for a long time. For example, a developer has spent a lot of time to enter, so there is a problem, that is, why do developers expect a longer working time for some patches? How long will they expect after they send the patch?

Andrew Morton: I think I am very concerned about regression testing. In fact, I don't want this patch to affect many people's systems if it is not good. If a running system is used by everyone, the impact is very large. Therefore, you must be very careful and cautious, make sure that there are no knowledgeable people to touch or modify the content further. Therefore, it takes a long time for me to understand this patch. This is why many patches have taken a long time, however, I am confident that it will not have a great impact, so I will decide to add it to the new version. Because this is a kernel that everyone is using, It is very serious.

Jonathan Corbet: I want to ask them questions in our audience, so I want to open a forum and ask them questions.

Questioner: we will launch a new Linux version recently. Why do we need this Tree? because many Linux contributions have caused many errors, we hope to remove the errors and defects, is that the reason? I have two questions. The first one is that more developers are involved in kernel development. Do you mean we will encounter more serious and chaotic problems in the future?

Andrew Morton: I am the first person to test every person's work. Every patch is tested. If this job is done by him, it can help me a lot. I believe they will test these jobs carefully and carefully, and they will be better off working with Steven and me. If we incorporate this Tree, we will test the results in the next year or two, so I believe we will spend a lot of time filtering out the patches that we didn't do before.

Questioner: Thank you very much.


Q: Linux always lags behind Windows. Is there a way to solve this problem? Let's release an updated version as soon as possible, which is the first problem. The second problem is power management. So far I think the power management system is not very good compared with Windows. What kind of work has the Linux community done to improve the power management system? Thank you.

Andrew Morton: It is very interesting that Intel was involved in this process. When they launched a new device, it was a Linux kernel, they actually started this work before the launch of these devices. They have done very well in this regard. Our current problems are for device manufacturers, they only obtained a 30% or 20% market share, and they were not fully involved. This is why LinuxDriver is slow. I think a perfect solution is to increase the market share of Linux, when the commercial value of LinuxDriver is higher, the LinuxDriver support will be better, and we hope to happen as soon as possible, but this does take some time. Indeed, Intel is very active in this field, whether it is Intel's hardware, and in general they are very active in kernel power management. In addition, when talking about power supply, I need to talk with some experts in charge of power supply. But at present, power management is becoming more and more stable, which is a problem area, but it is also an area that everyone is actively developing. I think power management may be different for different application systems. For example, there is a very general method for detecting how long it takes to play a DVD in a notebook, if you can finish the entire DVD, the power management is good.
There is a problem with DVD management, that is, hardware. If your hardware support is poor, it is also because of the small market share, therefore, the efficiency of power management is low. Thank you.

Questioner: You know that our current kernel is getting bigger and better. I think we must take some measures or some principles to ensure that our kernel is getting better and better, instead of developing in another direction, can you tell us what measures we can take or what principles we can apply? I don't understand your central idea. I am from the perspective of quality and functionality in the kernel development team, what measures and principles do you use to ensure better kernel functionality and quality?

Andrew Morton: there are many ways to measure and there is a tool that verifies a combination of code. Linux also has tools to help manage the quality of this aspect. Many of our QAA rely on those who download Linux from the Internet, in fact, there is a big difference between the Linux kernel and other software projects, that is, most of the Code cannot be tested by code writers. For example, some people on their own machines can develop a code, which is completely correct for them. However, for example, another person uses a different computer system to establish a file system on it, therefore, you cannot test a code on one or one hundred machines. You need tens of thousands of machines to test the Code together, therefore, it is very difficult to collect such a measurement evidence. This code has been introduced to hundreds of machines, causing problems. You can see the results of files in the past and have accumulated experience, but you can also have some systems. For example, let's see if the code suddenly disappears or what kind of interruption occurs. This is a very complicated problem. I have not seen any company or anyone put resources to solve this problem. Thank you.

Jonathan Corbet: I think we have time to answer another question, the last one.

Q: I have a question. We know that Linux is developing very fast, but some people complain that the API is changing too fast. Do you think this is a problem for some companies? Do you have any suggestions?

Andrew Morton: I think this API is indeed like this, so we do make regular adjustments now, but we are still very cautious in this regard. So I think what you said just now is the driver from the kernel to this kind of device, so we will change it often and it does bring a lot of trouble to many people, but we have nothing to compromise in this regard, because we want to improve our kernel performance, if we have to adjust this interface to improve our kernel, we have to make adjustments, because if we want to improve this interface without making any adjustments, then this kernel will not be able to move forward, this kernel will not be good, it will not be able to make progress, then this is necessary, this is the decision we made for them. For these drivers, their solution is to integrate the driver into the kernel to adapt to this change, in this case, there will be no major problems, so this is what I have discussed with many companies just now. It generally took year 34 years. For many companies, they always have this problem when they use Linux. I have been talking to them about how to solve this problem. They must integrate their code into the kernel, this problem can be solved to reduce their costs, but for every company, if you decide to enter Linux To adapt to this process. They often encounter quality problems, intellectual property problems, and so on. However, despite all kinds of painful problems, they generally solve them in the end. This is generally the case, but it usually takes year 34 to make the company really better.

Jonathan Corbet: I want to thank Andrew Morton for answering our questions just now.