When the Linux 5.8 Release Candidate opened for testing recently, the big news wasn’t so much what was in it, but itssize. As Linus Torvalds himselfnoted, “despite not really having any single thing that stands out … 5.8 looks to be one of our biggest releases of all time.”
True enough, RC 5.8 features over 14,000 non-merge commits, some 800,000 new lines of code, and added around a hundred new contributors. It might have gotten that large simply because few have been traveling thanks to COVID-19, and we’ve all been able to get more work done in a release window than usual. But from the perspective of this seasoned Linux kernel contributor and maintainer, what is particularly striking about the 5.8 RC release is that its unprecedented size just was not an issue for those that are maintaining it. That, I’d argue, is because Linux has the best workflow process of any software project in the world.
What does it mean to have the best workflow process? For me, it comes down to a set of basic rules that Linux kernel developers have established over time to allow them to produce relentlessly steady and reliable progress on a massive scale.
One key factor is git
It’s worth starting with a little Linux history. In the project’s early days (1991–2002), people simply sent patches directly to Linus. Then he began pulling in patches from sub-maintainers, and those people would be taking patches from others. It quickly became apparent that this couldn’t scale. Everything was too hard to keep track of, and the project was at constant risk of merging incompatible code.
That led Linus to explore various change management systems including BitKeeper, which took an unusually decentralized approach. Whereas other change management systems used a check-out/modify/check-in protocol, BitKeeper gave everyone a copy of the whole repo and allowed developers to send their changes up to be merged. Linux briefly adopted BitKeeper in 2002, but its status as a proprietary solution proved incompatible with the community’s belief in open source development, and the relationship ended in 2005. In response, Linus disappeared for a while and came back withgit, which took decentralized change management in a powerful new direction and was the first significant instantiation of the management process that makes Linux development work so well today.
Here are seven best practices — or fundamental tenets — that are key to the Linux kernel workflow:
Each commit must do only one thing
A central tenet of Linux is that all changes must be broken up into small steps. Every commit that you submit should do only one thing. That doesn’t mean every commit has to be small in size. A simple change to the API of a function that is used in a thousand files can make the change massive but is still acceptable as it is all part of performing one task. By always obeying this single injunction, you make it much easier to identify and isolate any change that turns out to be problematic. It also makes it easier for the patch reviewer only to need to worry about a single task that the patch accomplishes.
Commits cannot break the build
Not only should all changes be broken into the smallest possible increments, but they also can’t break the kernel. Every step needs to be fully functional and not cause regressions. This is why a change to a function’s prototype must also update every file that calls it, to prevent the build from breaking. So every step has to work as a standalone change, which brings us to the next point:
All code is bisectable
If a bug is discovered at some point, you need to know which change caused the problem. Essentially, a bisect is an operation that allows you to find the exact point in time where everything went wrong.
You do that by going to the middle of where the last known working commit exists, and the first commit known to be broken, and test the code at that point. If it works, you go forward to the next middle point. If it doesn’t, you go back to the middle point in the other direction. In that way, you can find the commit that breaks the code from tens of thousands of possible commits in just a dozen or so compiles/tests. Git even helps in automating this process with the git bisect functionality.
Importantly, this only works well if you abide by the previous rules: that each commit does just one thing. Otherwise, you would not know which of the many changes in the problem commit caused the issue. If a commit breaks the build or does not boot, and the bisect lands on that commit, you will not know which direction of the bisect to take. This means that you should never write a commit that depends on a future commit, like calling a function that doesn’t exist yet, or changing the parameters of a global function without changing all its callers in that same commit.
Never rebase a public repository
The Linux workflow process won’t allow you to rebase any public branch used by others. Once you rebase, the commits that were rebased will no longer match the same commits in the repositories based on that tree. A public tree that is not a leaf in a hierarchy of trees must not rebase. Otherwise, it will break the trees lower in the hierarchy. When a git repository is based on another tree, it builds on top of a commit in that tree. A rebase replaces commits, possibly removing a commit that other trees are based on.
Git gets merging right
Getting merging right is far from a given. Other change management systems are a nightmare to merge code from different branches. It often ends up with hard to figure out conflicts and takes a huge amount of manual work to resolve. Git was structured to do the job effortlessly, and Linux benefits directly as a result. It’s a huge part of why the size of the 5.8 release wasn’t really a big deal. The 5.8-rc1 release averaged 200 commits a day, with 880 total merges from 5.7. Some maintainers noticed a bit more of a workload, but nothing was too stressful or would cause burnout.
Keep well-defined commit logs
Unfortunately, this may be one of the most essential best practices that are skipped over by many other projects. Every commit needs to be a stand-alone, and that includes its commit log. Everything required to understand the change being made must be explained in the change’s commit log. I found that some of my most lengthy and descriptive changelogs were for single line commits. That’s because a single line change may be for a very subtle bug fix, and that bug fix should be thoroughly described in the changelog.
A couple of years after submitting a change, it is highly unlikely that anyone would know why that change was made. A git blamecan show what commits changed the code of a file. Some of these commits may be very old. Perhaps you need to remove a lock, or make a change to some code and do not precisely know why it exists. A well-written changelog for that code change can help determine if that code can be removed or how it can be modified. There have been several times I was glad I wrote detailed changelogs on code as I had to remove code, and the changelog description let me know that my changes were fine to make.
Run continuous testing and integration
Finally, an essential practice is running continuous testing and continuous integration. I test every one of my pull requests before I send them upstream. We also have a repro called linux-next that pulls in all the changes that maintainers have on a specific branch of their repositories and tests them to assure that they integrate correctly. Effectively, linux-next runs a testable branch of the entire kernel that is destined for the next release. This is a public repo so anyone can test it, which happens pretty often – people now even release bug reports on code that’s in linux-next. But the upshot is that code that’s been in linux-next for a couple of weeks is very likely to be good to go into mainline.
Best practices exemplified
All of these practices enable the Linux community to release incredibly reliable code on a regular 9-week schedule at such a massive scale (average of 10,000 commits per release, and over 14,000 for the last release).
I’d point to one more factor that’s been key to our success: culture. There’s a culture of continuous improvement within the kernel community that led us to adopt these practices in the first place. But we also have a culture of trust. We have a clear pathway via which people can make contributions and demonstrate over time that they are both willing and able to move the project forward. That builds a web of trusted relationships that have been key to the project’s long term success.
At the kernel layer, we have no choice but to follow these practices. All other applications run on top of the kernel. Any performance problem or bug in the kernel becomes a performance problem or bug for the applications on top. All error paths must exit peacefully; otherwise, the entire system will be compromised. We care about every error because the stakes are so high, but this mindset will serve any software project well.
Applications can have the luxury of merely crashing due to a bug. It will annoy users, but the stakes are not as high. Quality software should not take bugs lightly. This is why the Linux development workflow is considered the golden standard to follow.
From its efforts to reshape computing through open source to its aggressive push to increase internet connectivity around the world, Facebook is a leader in open innovation. Perhaps more important today than ever, Facebook’s focus on democratizing access to technology enhances opportunity and scale for individuals and businesses alike. That’s why we’re so excited to announce the company is joining the Linux Foundation at the highest level.
Facebook’s sponsorship of open innovation through the Linux Foundation will help support the largest shared technology investment in history with an estimated $16B in development costs of the world’s 100+ leading open source projects and supports those project communities through governance, events and education. The company is also already the lead contributor of many Linux Foundation-hosted projects, such as Presto, GraphQL, Osquery and ONNX. It has been an active participant in Linux kernel development, employing key developers and maintainers across major kernel subsystems.
In addition to these efforts, Facebook has a long history of leveraging open source to unlock the potential of open innovation:
Facebook created a unique dataset of over 100,000 videos and launched the Deepfake Detection Challenge in order to accelerate development of new ways to detect deepfake videos. This open, collaborative effort will help the industry and society at large meet the challenge presented by deepfake technology and help everyone better assess the legitimacy of content they see online.
Facebook’s Data for Good program enables geographic data to be shared with the aim of addressing some of the world’s greatest humanitarian issues, including COVID-19.
Facebook also leads the industry in open hardware, having founded the Open Compute Project (OCP), which uses open source to enable the creation of efficient, flexible, and scalable hardware designs for data centers.
By creating and sustaining an open source ecosystem around PyTorch, Facebook also accelerates the pace at which data scientists and developers can leverage the power of artificial intelligence and machine learning in computer vision, natural language processing, and other disciplines.
Facebook’s React.js library powers some of the world’s most popular websites and has become the standard for frontend web development due to its simplicity and flexibility.
Facebook’s commitment to the open source community can be seen in both its multi-million dollar investments and its genuine passion for technology development. It is this combination that makes the company an incredible supporter of the open source developer community.
As a Platinum member of the Linux Foundation, Facebook’s Kathy Kam joins the LF board. Kathy is head of Open Source at Facebook where she manages the Open Source Engineering, Developer Advocacy, and Open Source Program Management teams. Kathy is a 20-year engineering, product management, and developer relations leader previously with Google and Microsoft.
The Linux Foundation would like to reiterate its statements and analysis of the application of US Export Control regulations to public, open collaboration projects (e.g. open source software, open standards, open hardware, and open data) and the importance of open collaboration in the successful, global development of the world’s most important technologies. At this time, we have no information to believe recent Executive Orders regarding WeChat and TikTok will impact our analysis for open source collaboration. Our members and other participants in our project communities, which span many countries, are clear that they desire to continue collaborating with their peers around the world.
As a reminder, we would like to point anyone with questions to our prior blog post on US export regulations, which also links to our more detailed analysis of the topic. Both are available in English and Simplified Chinese for the convenience of our audiences.
The Linux Foundation has partnered with edX to update the Open Source Jobs Report, which was last produced in 2018. The report examines the latest trends in open source careers, which skills are in demand, what motivates open source job seekers, and how employers can attract and retain top talent. In the age of COVID-19, this data will be especially insightful both for companies looking to hire more open source talent, as well as individuals looking to advance or change careers.
The report is anchored by two surveys, one of which explores what hiring managers are looking for in employees, and one focused on what motivates open source professionals. Ten respondents to each survey will be randomly selected to receive a US$100 gift card to a leading online retailer as a thank you for participating!
All those working with open source technology, or hiring folks who do, are encouraged to share your thoughts and experiences. The surveys take around 10 minutes to complete, and all data is collected anonymously. Links to the surveys are at the top and bottom of this post.
Linux Foundation Executive Director Jim Zemlin recently spoke at Cloud Native + Open Source Virtual Summit China 2020. We’d now like to republish his opening comments and a guide on how to get involved with the TARS project, the open source microservices framework.
The pandemic has thrown our global society into a health and economic crisis. It seems like there are conflicts every day from all over the world. Today, I want to remind you that open source is one of the great movements where collaboration, working together, and getting along is the essence of what we do.
Open source is not a zero-sum game, but it has had an incredible impact on us in a net positive way. I like to remind everyone that open source is public goods that will be freely available to everyone worldwide, no matter what wind of political or economic change brings us. The LF is dedicated to all of that.
Today, we are working hard to help folks during hard times, expanding our mentorship programs with over a quarter of million dollars of new donations to allow people to come in and train themselves on new skills during this tough time. We had a wonderful set of virtual events with thousands of people from hundreds of companies from countries worldwide working together.
We want to bring the power of open source to help during these times and have several new initiatives that we are working on. Most notably, our recently launched LFPH initiative, which has started with seven members: Cisco, doc.ai, Geometer, IBM, NearForm, Tencent, and VMware, and it’s hosting exposure notification projects such as Covid-Shield and Covid-Green, which are currently being deployed in Canada, Ireland, and several U.S. states to help find and reduce the spread of COVID-19.
We are also working on a considerable number of new initiatives, which I will talk about. Still, I like to remind you of what we are here to talk about, which is cloud computing, and how much cloud computing has impacted all of us. Microservices are an essential part of that. In China we are seeing the TARS project; the microservices framework is really taking off.
Two years ago, TARS joined the Linux Foundation, and ever since its community has been growing and new projects and contributors have been coming in. The TARS project provides a mature, high-performance microservices framework that supports multiple programming languages. We will talk more about the TARS Foundation in a little bit, but the microservices ecosystem has been growing and quickly turning applications and ideas in scale.
In addition to TARS, we have been seeing amazing work going on in the open source community. It begins with things such as the Software Package Data Exchange specification (SPDX), which was recently contributed as an international specification to the ISO/IEC JTC 1 for approval. This will help us track the usages of open source software across a complex global supply chain and reaffirm our commitment to the global movement.
We also see growth and projects with recent releases, such as our networking project, the Open network automation platform Frankfurt release, which is being used to automate the networks and edge computing service for telecommunication providers, cloud providers, and enterprises.
We‘ve seen new projects join our organization. One good example is MLflow — this project was contributed to our organization from Data Brick. This project has had an impressive community with over 200 contributors, which has been downloaded more than 2 million times. MLflow is part of the LF AI initiative. It will be a neutral home and open governance model to broaden the adaptation and contribution of things like MLflow. We have also seen new projects come to our organization, such as the FinOps Foundation, the consortium of financial companies. We are working together to grow the use of open source throughout our global financial system.
It’s impressive to see all the different projects that have been coming. And today, I’d like to introduce the TARS Foundation formally. TARS has been an amazing project, and in just the last few years, I’ve noticed that developers here in China are for the first time incubating and sharing new open-source projects in China and the rest of the world.
And the rest of the world is watching the progress of open source projects and seeing fantastic work. We are so proud of the work that is coming out of TARS.
You know, Just like the Linux Foundation is about more than Linux, the TARS Foundation is more than just TARS. It’s a microservices ecosystem.
Unfortunately, because of the COVID-19 pandemic, we had to cancel the Linux Foundation Member Summit this Spring, and we were unable to announce the TARS Foundation at that time.
But today, the Linux Foundation is proud to announce again that the TARS project has become the TARS Foundation, an open-source microservice foundation within the overall framework of the Linux Foundation, and its outcome has been rapid growth for both the TARS project and projects associated with TARS. TARS has really taken off, and it’s just amazing to see the amount of development.
We hope the TARS foundation will create a neutral home for additional projects for solving significant problems surrounding microservices, including but not limited to:
Agile development, DevOps best practices, and the comprehensive governance that we have will enable multi-languages, high performance, scalable solutions.
It is my pleasure to present what the TARS Foundation has achieved in the open source community.
There are many companies whose contributions are instrumental in establishing TARS’ microservices ecosystem. The TARS Foundation is proof of that. Currently, the TARS Foundation has Arm and Tencent as premier members and five general members: AfterShip, Ampere, API7, Kong, and Zenlayer.
In terms of TARS applications, it serves more than 100 companies from different industries, including Edge, E-sport, Fintech, Streaming, E-commerce, Entertainment, Telecommunication, Education, and more.
Furthermore, the TARS Foundation is striving to expand its microservices ecosystem, and it’s incorporating more functions such as Testing, Gateway, and Edge, to name a few. So far, the TARS Foundation has more than 30 projects.
Developers around the world are starting to realize that the TARS project is amazing and contribute as such. There are 12,000 developers actively using TARS. Also, 150 developers contribute code to TARS projects, from companies like Arm, Tencent, Google, Microsoft, Vmware, Webank, TAL, China Literature, iFlytek, Longtu Game, and many more.
An overview of the TARS framework and how you can contribute to the open source microservices community
What is TARS?
TARS is a new generation distributed microservice application framework that was created in 2008. It provides developers and enterprises with a complete set of solutions to build, release, deploy, and maintain stable and reliable applications that run at scale.
“a neutral home for open source microservices projects that empower any industry to quickly turn ideas into applications at scale”.
The TARS Foundation’s goal is to address the most common problems related to microservices application, including solving multi-programming language interoperability issues, mitigating transfer issues, maintaining data storage consistency, and ensuring high performance while supporting a growing number of requests.
Many companies have successfully used TARS framework from diverse industries such as fintech, esports, edge computing, online streaming, e-commerce, and education, to name a few.
Here is a complete timeline of the TARS Foundation’s development:
Initially developed by Tencent, the world’s largest online gaming company, the TARS project has created an open source microservices platform for modern enterprises to realize innovative ideas quickly with the user-friendly technology in the TARS framework.
In March 2020, the TARS project transitioned into the TARS Foundation under the Linux Foundation umbrella, aiming to support microservices development through DevOps best practices, comprehensive service governance, high-performance data transfer, storage scalability with massive data requests, and built-in cross-language interoperability. TARS has a mission to support the rapid growth of contributions and membership for a community focused on building a robust microservices platform.
The TARS Foundation provides a great platform for developers who are interested in contributing to an open source project. The organization extends different opportunities for developers to contribute to open source projects and the possibility to take on leadership roles and create major contributions in the broader open source community.
There are Contributor, Committer, Maintainer, and Ambassador roles in their open source ecosystem, each having different requirements and responsibilities.
How to become a Contributor
To get involved with TARS open source projects, you can first become a Contributor by participating in software construction and having at least one pull request merged into the source code.
There are several ways for software developers to engage with the TARS community and become contributors:
Help other users and answer questions.
Submit meaningful issues.
Use TARS projects in production to increase testing scenarios.
Improve technical documentations.
Publish articles on applications and case studies related to TARS projects.
Make changes to the code and test it on your local machine.
Commit those changes.
Push the committed code to GitHub.
Open a new pull request to submit your changes for review.
Your changes will be merged into the master branch if accepted.
Now you did it! You’ve become a TARS Contributor, and you will receive a Contributor t-shirt!
How to become a Committer
A Committer is a contributor who has made distinct contributions to the TARS repositories and has accomplished at least one essential construction project or has repaired critical bugs. He or she can also take on some leadership opportunities.
The Committer is expected to:
Display excellent ability to make technical decisions.
Have successfully submitted and merged five pull requests.
Have contributed to the improvement of project code quality and performance.
Have implemented significant features or fixed major bugs.
After meeting the above requirements, you can submit a Committer request:
STEP 1: Provide your proof of the above criteria under Repo ISSUE.
STEP 2: Submit your pull request after you receive a response with instructions
STEP 3: Once your application is accepted, you will become a TARS Committer!
As a Committer, you are able to:
Control the code quality as a whole.
Respond to the pull requests submitted by the community.
Mentor contributors to promote collaborations in the open source community.
Attend regular meetings for committers.
Know about project updates and trends in advance.
How to become a Maintainer
Maintainers are responsible for devising the subprojects in the TARS community. They will take the lead to make decisions associated with project development while holding power to merge branches. They should demonstrate excellent judgment and a sense of responsibility for subprojects’ well-being, as they need to define or approve design strategies suitable for developing subprojects.
The Maintainer is expected to:
Have a firm grasp of TARS technology.
Be proactive in organizing technical seminars and put forward construction projects.
Be able to handle more complicated problems in coding.
Get unanimously approved by the technical support committee (TSC).
As a Maintainer, you have the right to:
Devise and decide the top-level technical design of subprojects.
Define the technical direction and priority of sub-projects.
Participate in version releases and ensure code quality.
Guide Contributors and Committers to promote collaborations in the open source community.
How to become an Ambassador
Passionate about open source technology and community, Ambassadors promote and support extensive use of TARS technology to a wider audience of software developers. Ambassadors’ expertise and involvement in TARS projects will also acquire greater recognition in the community.
The Ambassador can:
Become a general member of the TARS Foundation.
Participate in TARS Foundation’s projects as a contributor, lecturer, or blogger.
Engage with developers by presenting at community events or sharing technology articles on online media platforms.
Ultimately, the TARS Foundation encourages a contributor to becoming a member of the governing board and the Technical Support Committee (TSC). At this level, you will focus on the organization’s strategic directions and decision-making as a whole.
Contributing to open source projects has many benefits. It strengthens your development skills, and your code is reviewed by other developers who can give a new perspective. You are also making new connections and even lifelong friendships with like-minded developers in the process of contributing. This is the open source model that has built many tech innovations that all of us enjoy today. Its sustainability depends on a free exchange of ideas and technology in our global community. Open source value and innovations are embedded in developers like you who can attempt development challenges and share insights with the broader community.