domeet webmaster

At the same time, the different strategies and competing relationships between start-up teams and technology giants in their respective open source commercialization processes have given me a deeper understanding.

Undoubtedly, IP (Intellectual Property, Intellectual Property) has created two generations of great technology companies and high-barrier business models: p>

• Intel and Qualcomm that build monopoly positions with chip sales and baseband authorization;

  • Microsoft, which licenses PC makers with software, and Apple, which bundles and sells native operating systems and hardware directly.

    The developer-centric software and App ecosystem created by the latter two companies, coupled with the network effects of the software itself, have directly promoted the development of the Internet, cloud computing, big data and today’s AI and other technologies. In this process, the dependence of users and developers on the platform once again strengthened the ecological barriers.

    Barriers bring monopolies, which generate high profits.

    Behind all the technology, you can also see the shadow of another giant, Google, whose strategy seems to be the exact opposite of all the companies above. From open source Chrome browser, Android system, to deep learning framework Tensorflow and other technologies, before A16Z shouted “open source is consuming software” last year, I have firmly believed that Google is the biggest contributor and profit of software technology development in the past decade Or, open source is one of its most important business strategies.

    But after researching this time, when I saw the answer on Quora again, I gradually understood a basic logic:

    Selecting open source is essentially a technical direction issue, not a businessIndustry issues cannot be a business model.

    “Contemporary open source is to let more people use our framework to find more algorithms and deployment solutions that can be used in the industry.” Xie Yinan, vice president of contempt, told me that this is a deep learning framework “Tianyuan “The main purpose of Alpha is open source.

    The technical difficulty lies in first letting more than 1,400 R & D personnel inside the company to use it in their daily work based on unified technical standards. Secondly, they need to internalize the algorithms and models used in the industry to further feed back the algorithms behind the framework. Finally, the interface is encapsulated and standardized, and veterans and researchers who are used to Tensorflow and PyTorch can get started quickly with the help of technical documents.

    Going back to the example of the Android system, Google open sourced only the AOSP part of the year (Android Open Source Project) , the fee is GMS (Google Mobile Service, Google Mobile Service) , GMS is Google’s various applications and APIs, including maps, emails, Youtube and Application store, etc.

    Because Android’s kernel itself is based on Linux, the latter requires that Android’s core code must be freely available. GMS is Google’s “money printer”. Through the network effect of the software and the huge user base, mobile phone manufacturers eventually had to use it as a factory standard, so that Google continued the “high profit, closed source and The “exclusive” business model has established its dominance in mobile operating systems.

    So, from a technical perspective, it involves the underlying architecture, operating performance, user interface, and other parts of the operating system. Any developer or hardware manufacturer can conduct secondary development based on open source code. On the basis of complying with relevant open source licenses, secondary developers have the right to freely choose whether to open source,It depends on whether they want to profit directly from it.

    In addition, one year after Google open sourced Tensorflow, the performance of more of its products has been rapidly improved, and its own AI company Deepmind has subsequently announced the use of a new generation of Tensorflow as the underlying algorithm framework. It can be seen that open source is an opportunity for most companies to openly test the core technology’s leadership and availability.

    HashiCorp, a company that develops open-source tools to support multi-cloud deployments, just announced this month that it has completed a $ 175 million Series E financing. CTO Armon said in a discussion about open source: When faced with the choice of open source or closed source, the team will first identify the problem is from “technical complexity ( technical complexity) “still involves” organizational structure (organizational) “. If it affects the basic use of the tool or product, this is a technical problem and must be open source after being solved; if it comes from an island or efficiency issue within the company, it does not need to be disclosed.

    Second, if you want to make an open source project (project) into a successful open source product (product) , this is a business issue.

    There is an open source project that has not been closed in front of users since its launch. The team consists of several doctoral students from UC Berkeley. The main reason for the rejection was that users were worried that the team might dissolve after graduation. So in 2013, everyone decided to invest full time and set up a company called Databricks. Now they have completed the F round of financing with a valuation of more than 6 billion US dollars. In the first three years, the team has only done one thing, promoted the parallel computing framework Spark project and actively maintained the developer community. It wasn’t until 2015 when demand erupted suddenly that code contributions surged that they started thinking about how to commercialize.

    This is a process that most teams will go through. More interestingly, Databricks CEO Ali and HashiCorp’s Armon both believe that it is often a streamlined team or company that can provide core support behind a successful open source project. , Up to two.

    “When you think about why you want to open source, then you need to figure out how to open source, including version iteration and scene landing.” This is Xie Yinan’s first recommendation to entrepreneurs. Despise has made a clear plan for the open source product line, from the supported CPU types to the coverage of multiple embedded devices.

    (Source: Contempt Open Source Conference)

    In addition, the commercialization of start-up teams (or companies in growth stage) Compared with large companies, there are mainly the following differences, FYI:

    1. Different starting points: Large companies may open source early in a project. With their appeal, they hope that more people “contribute” to iteration together. The startup team will open again when the product is relatively mature, hoping to attract users as soon as possible. In-depth “use”, focusing on improving the comprehensive performance of products in the industrial environment.

    “Open source is not my impulse, but it is thoughtful and long-planned.” Tao Jianhui, CEO of Taosi Data, an open source big data platform for the Internet of Things, wrote on his account. Three years to write code, less than ten R & D teams, and the project got more than 10,000 stars in just three months after launching on GitHub, which is very difficult for startups;

    2. The strategic significance is different: whether the product or the ecology may be only one part of the business strategy of the large company, and it is all for the startup company’s products and users.

    The open source database listed company MongoDB attaches great importance to the self-service open source product “Community Server”, and considers it to be the company’s most important sales funnel. When users in the open source community want to build applications on the database, they can try the hosted DbaaS for free (Database as a Service, Database as a Service) ProductionAtlas, you will need to pay when the usage is further increased. The company’s sales will continue to screen sales targets with higher-level needs and payment capabilities among the paying user groups, and provide them with enterprise-level products Enterprise Advanced, including value-added services such as dedicated servers and operation and maintenance tools;

    3. Different operating strategies: Therefore, large companies establish “span classes” (partnership) , including traditional software giants, consulting companies, ISV class = “text-remarks” label = “Remarks”> (Independent Software Developers) , integrators, SaaS service providers and other partners, while the startup team pays more attention to maintaining the “Developer Community” -remarks “label =” Remarks “> (community) .

    Especially for the Chinese team, we should pay attention to the issue of internationalization from the beginning, “from English documents, tutorials, cases to tutorial preparation, establish a strict code review process, and value the contribution of each user”, Yuan Jinhui, the CEO of first-class technology, told me that his company’s deep learning framework, OneFlow, will also be open source in a few months.

    So, when you have enough thoughts and preparations, and you decide to devote yourself to commercial operation, where is “Dongfeng”?

    Third, find enterprise customers. The focus of “SaaS” is not “Software”, but “Service”, just as “Cloud service” is not about cloud, but service.

    Although start-up teams and big companies have different starting points and strategies in the early days of open source, the boundaries of these behaviors will gradually blur after the development of the former or after being acquired by big companies, and new differences will soon appear:

    • In the past, most users were “developers”, and now developers cannot fully represent his “business” , The real and complete needs of enterprise customers must be re-investigated;

      • In the past, the user found that the code submission was “pull request” and waited for your reply. Now the customer will call your customer service in the middle of the night and ask to deal with the problem immediately; / ul>

        • Customers used to download software independently and deploy it privately. I hope you provide on-site support and maintenance if necessary. Now they are used to putting more applications and data in the cloud, and hope you can provide similar “services” and charging methods.

          You will find that the emergence of these new features has nothing to do with the open source or not.

          The era of open source 1.0 led by RedHat came to an end with RedHat being acquired by IBM for $ 34 billion in 2018. The “Dongfeng” of the 2.0 era is a new business model, product route and application in the face of new features. Architecture-SaaS, “Developer-led (Developer-oriented) ” and cloud native.

          I explained the importance and opportunities of SaaS and cloud-native in the opportunity of the rise of the platform after Salesforce. Here are a few more about the “Developer Orientation”:

          • First in the early days of open source, it’s important to get advocacy from developers in the community. The reason is No more details;

            • After productization, we need to provide additional services for developers so that they can focus on the application level Development. With the increasing influence of developers’ purchasing decisions in the company, the product will likely form a sales conversion for enterprise customers. This is similar to the reason that Zoom can change the company’s top-down procurement model, precisely because employees’ influence in corporate IT procurement has increased year by year;

              • Finally, deepen the research and development of enterprise customers and even business processes to build a full-stack solution. This may be more important for companies engaged in related open source projects at the infrastructure level, for reasons explained later.

                
                

                Tao Jianhui also shared, “From the beginning of the design, it was decided to create a full-stack time-series data processing tool, not only a time-series database, but also a series of functions such as caching, streaming computing, message queues, and subscriptions. In order to minimize the product’s consumption of system resources and the complexity of maintenance.

                Only by being developer-oriented can we finally be customer-centric.

                Finally, in the face of the inevitable “If BAT (FAANG abroad) , what do you do?” This Problem, choose an open source license carefully and protect the IP.

                A quick summary of the types of open source licenses:

                (Source: https://www.cnblogs.com/newcaoguo/p/7103249.html)

                What prompted me to think about this issue came from the following events:

                • The original GPL protocol, due to the network service (Web service) The rise of the company (such as Google) has created certain vulnerabilities. For example, using open source software under the GPL protocol, which is not released on the network but only provides services through the cloud, then the company is free to Use GPL protocol but not open source your own private solution. Therefore, the emergence of AGPL is to make up for this loophole;

                  • In October 2018, MongoDB announced its open source license switch from AGPL v3 to its own defined SSPL (Server Side Public License) . SSPL will explicitly require the cloud service provider hosting the MongoDB instance (especially for Asian companies, your details) either obtain a commercial license from MongoDB, or Open source its service code to the community in order to fight against evasion of AGPL supervision;

                    • In January 2019, Amazon AWS launched Managed Document Database Service DocumentDB, which is fully compatible with MongoDB. The official website claims to “provide the performance, scalability, and availability needed to run mission-critical MongoDB workloads at large scale”, while proposing to help users “easily migrate MongoDB databases on-premises or on the Amazon Cloud to DocumentDB and Hardly any downtime. ” Immediately after the launch of the product, it caused huge controversy in the open source community, and MongoDB’s stock price fell 13% on the day.

                      AGPL tried to make up for the above problem of “Web service loophole (Network Service Vulnerability) “, but it was known by launching SSPL from MongoDB not effectively. What worries me even more is that the emergence of DocumentDB pushed a technical direction issue directly to the level of commercial feasibility, and made “cloud services” a double-edged sword in front of many startups that commercialized open source with the SaaS model: / p>

                      Customers hand over data and calculations to open source companies, and the quality of service of open source companies (including performance, scalability, and availability) depends on (AWS, Alibaba, Microsoft, etc.) at the same time, their code is also exposed to the public environment. So whether in terms of economies of scale or service quality, the latter naturally has a certain advantage in the infrastructure layer.

                      Of course, as mentioned above, with the emergence of special needs such as full-stack services, hybrid cloud deployment, and differentiated processing on IoT devices and edges, the design, development, and maintenance of related products on the cloud were initially Startups will gradually form their own moats.

                      So, when the company chooses SaaS as the basic mode of business operation, it will judge whether the moat is strong or not and its commercial value. In the end, it will return to the core indicators of SaaS: such as growth, churn (Churn rate) and income retention rate (NDR) , etc.

                      In the three-element AI algorithm, computing power, and data, Google has strong advantages over the latter two. After tasting the sweetness on open source Android, the open source deep learning framework Tensorflow became logical in 2015. Baidu also open-sourced its deep learning framework “Flying Paddle” in 2016. Last week Tsinghua University officially opened its self-developed deep learning framework “Plan”, and this week’s deficient “Tianyuan”. There are rumors that Huawei also plans to open MindSpore, a deep learning platform that has already been launched this year. Take a closer look. Each has its own technical characteristics and business abacus.

                      In the future, we will definitely see more domestic open source projects and technology giants’ actions in frontier areas such as AI, infrastructure, and IoT.

                      But it needs to be very careful that this is not comparable to the so-called “domestic substitution”. The reasons are as follows:

                      • The open source community is a place where global developers compete and share knowledge. Developer Support “is neither persuasive nor influential enough;

                        • From “developer-oriented” to “customer demand-centric”, except for some sensitive areas The customer base of open source companies should be regardless of country. Moreover, if the security and privacy of enterprise customers cannot be guaranteed, how to ensure security in those high-risk areas?

                          • Open source products with SaaS as the main model, due to self-service (Self-service) will naturally bring global reputation to the company. This will directly affect the core indicators of SaaS, and any protective policy will not benefit the company’s long-term business value.

                            Even for all “domestic substitution” projects, I agree with Guo He, Managing Director of Songhe Capital, in ” Boiling Ten Years: An Irreversible Tide of Domestic Alternatives “ Views:

                            “What we want is a real replacement, not a great spare.”

                            So, what kind of projects and companies will we see in this era of “open source is consuming software”?

                            A long-lasting open source service on the front end is bound to have a development team with a high degree of cohesion and a global vision, as well as a highly commercialized company formed around this team, regardless of country.

                            This article comes from WeChat public account: I think about the pot and I’m at (ID: angelplusdevil) , author: I think I pot in GN