Author | David Rosenthal
Translator | Crescent Moon, Editor in Charge | Zheng Liyuan
Head picture | CSDN download from Visual China
Exhibit | CSDN (ID: CSDNnews)
The following is the translation:
Recently, Michael Dorner, Maximilian Capraro and Ann Barcomb jointly published a paper "Quo Vadis, Open Source? The Limits of Open Source Growth" (Address: https://arxiv.org/pdf/2008.07753.pdf), which is presented in the paper. I posted a piece of statistical data and discussed the substantive issues affecting the limitations of open source development.
Here are some excerpts:
Open source software plays an important role in the software industry. Previous research has shown that open source is growing exponentially, and this growth trend is even close to exponential. However, when resources are not unlimited, this growth cannot continue indefinitely. In this study, we took more than 224,000 open source projects in the past 25 years as the research object, and carried out four cumulative measurements of the scale and growth of open source. We measured the number of lines of code, number of submissions, contributors, and changes in the life cycle status of these projects over time, and finally came up with three research results, which are widely cited. We found that since 2016, the number of active open source projects has been declining, and the number of contributors and submissions has been declining since the peak in 2013. Although open source projects used to grow exponentially at first, they are no longer growing. We believe that open source has reached saturation.
As shown in the figure below, the authors of the paper observed that the monthly submission rate increased exponentially until 2010, peaked in 2013, and then declined all the way until 2019 and 2007 were the same.
I think this is in line with W. Brian Arthur's technical and economic model. When a new type of niche market has just emerged, people will put a lot of effort into it. However, as time goes by, the increase in returns to scale will push a few, perhaps only one, to dominate this market.
Those projects that can quickly add features (submit) will grow rapidly and gradually occupy this market. The losers can't keep up, so their submission rate will drop, gradually become less and less active, until they are finally abandoned. As the competition for the winners decreases, the features they add will also decrease. Their submissions are increasingly focused on fixing bugs or resolving vulnerabilities, so the incidence of bugs will also decrease, and eventually stabilize at a slower and constant rate. Please note that this model is not suitable for a few large aggregate projects, such as the Linux kernel.
The picture below is the most interesting, it shows the number of people at different stages of the life cycle of an open source project. The author of the original paper observed:
We can confirm that as of 2013, all projects have grown exponentially, but most of the projects are inactive, and often one submission is not received in a certain month...Most of the inactive projects are not available Funding, they were all abandoned.
Beside this there is:
The proportion of actively developed open source projects (that is, projects that receive at least one grant every month) is small and remains roughly the same over time.
If I'm not mistaken, these active projects include those large projects, usually infrastructure, and some smaller projects, which are still competing for their respective markets. If you group the activity data of Dorner et al. according to age and scale, you may get two groups:
● Larger, older, and actively maintained projects.
● Smaller, newer projects, these projects gradually become inactive after a period of time.
Please note that smaller inactive projects may already be widely used. It has occupied its own market, has all the important functions, and fixes all easy-to-find bugs. A small number of small projects may remain in this mature state until they are eliminated.
In addition, the author of the original paper also observed that the number of monthly active contributors behaves similarly to the number of monthly submissions, as shown in the figure below. Until 2010, the number of contributors has been increasing exponentially, peaking in 2013, and then declining, until 2019 its growth rate was the same as 2017.
Of course, if the productivity of all contributors is the same, then this statistic also meets people's expectations. However, as we all know, programmer productivity has a long-tailed distribution. However, assuming that this distribution does not change significantly over time, the difference between the mean and the median of productivity is relatively stable, so the result is the same.
The authors of the paper gave the following explanations for the open source saturation phenomenon they observed:
● The number of developers willing to contribute code has decreased, while paid development work has not increased accordingly.
● Due to the company's resource management, many spontaneous contributions have turned into paid contributions, resulting in a reduction in the effective time spent by each contributor.
● More and more projects require continuous participation, so more and more people tend to reduce their spontaneous contributions.
● The two-generational transition from collective voluntary behavior to self-interested voluntary behavior (the average age of contributors was 31 in 2005 and 30 in 2017), which may be due to the increasing positive effects of participation in open source on career development result.
● The increase in code complexity has led to fewer and fewer developers with corresponding technologies, and discourages newcomers.
● The degree of formalization of software projects is getting higher and higher, and developers need to make a lot of effort to comply with submission or basic guidelines.
● The quality of contributions has decreased, which has led to a decrease in acceptance rate and an overwhelming burden on reviewers and submitters.
These reasons seem reasonable, but I would like to add one more based on W. Brian Arthur's model. As time goes by, it is more and more likely that the new field will be closer to the field of the existing winner. Therefore, the demand for new features is more likely to be met by one of the regular contributors of an existing project adding a small number of commits to the existing project, rather than a large number of commits from several new contributors to start a new project. This is like the lack of antitrust, so the technology oligarch suppresses competition from startups.
Finally, I want to thank Glyn Moody, who concluded:
This new study shows that the open source community that has been selflessly dedicated for decades is showing signs of altruistic fatigue. Companies should reflect on their feedback to open source projects and make unprecedented efforts.
Original: https://blog.dshr.org/2020/09/open-source-saturation.html
This article is a CSDN translation, please indicate the source of reprint.
更多精彩推荐
☞CSDN 创始人蒋涛：选择长沙作“大本营”，打造开发者中心城市
☞这都是啥软件？你能猜到吗？| 每日趣闻
☞杜甫在线演唱《奇迹再现》、兵马俑真人还原……用AI技术打破次元壁的大谷来参加腾讯全球数字生态大会啦！
☞开放源码，华为鸿蒙HarmonyOS 2.0来了
☞20张图，带你搞懂高并发中的线程与线程池！
☞跨链，该怎么跨？
点分享点点赞点在看