johnlui/PPHC: A Free Chinese Book on High-Concurrency Architecture, Split Into Five Parts

What PPHC Is and Who It Is Written For

PPHC is not software. The repository johnlui/PPHC contains a book, 《高并发的哲学原理》, written by Lü Wenhan, and the README states the work is open source under CC BY-NC-ND 4.0 and free to read for GitHub users. The primary language listed for the repository is Rust, which is misleading if you arrive expecting a high-concurrency framework in Rust; the repository layout at top level is five part directories, a LICENSE file, an OldArticles directory, README.md and an assets directory. The Rust label reflects nothing about the content.

The stated goal is narrow and worth quoting: the book wants to discuss one philosophical principle behind high-concurrency problems, "找出单点，进行拆分", find the single point and split it. Everything else in the text is an application of that idea. The README says the open-source edition has 12 chapters, 83 articles and 167,547 characters in total, which puts it in book territory rather than blog-post territory.

The intended reader is a backend or infrastructure engineer who already knows what a load balancer is and wants the layer underneath. The author's own credentials are listed in the README: maintaining a monolithic CMS that handled two million PV per day in 2017, optimising a single-machine PHP shop to over QPS 1000 in 2020, and designing a distributed flash-sale system that reached a peak of five million GMV per minute at over QPS 10000 in 2021. Those are the numbers the README gives, and they set the register: practitioner notes, not academic treatment.

The Split-the-Single-Point Argument That Runs Through the Book

The mechanism of the book is its ordering. It starts from static and dynamic separation, then moves through Apache, Nginx, epoll, virtual machines, Kubernetes, asynchronous non-blocking I/O, coroutines, application gateways, L4 and L7 load balancers, routers, switches, LVS, software-defined networking, Keepalived, DPDK, ECMP, fully redundant architecture, user-space network cards, centralised storage, distributed storage, PCIe 5.0, CXL, InnoDB's three-level index, in-memory caches, KV databases, column stores, in-memory databases, Shared-Nothing, compute-storage separation, Paxos, microservices, peak shaving and geographic splitting.

That list is the argument. Each item is a place where a single point can be identified and then divided, whether the single point is a process, a machine, a network path, a disk, or a consensus participant. The README frames the endpoint as designing an architecture that could serve all of humanity, which is a rhetorical flourish rather than an engineering claim, and readers should treat it as framing.

The data flow of the repository is equally simple: Markdown files under part directories, images under assets, older material under OldArticles. There is no build step, no server, no configuration. If you clone it, you get text. The homepage at pphc.lvwenhan.com serves the same content in readable form, and the README notes that the PDF download link is in the top right corner of that site.

Reading the Book Locally: Clone, Browse, and Get the PDF

There is no package to install. The repository is a text collection, so the first real use is cloning it and reading the part directories in order. The README points readers at the hosted version and at the PDF link on that site, and gives no local build instructions, so the commands below are limited to what the repository layout supports.

Clone the repository and list the top-level entries. You should see the five part directories that hold the book text, alongside LICENSE, README.md, OldArticles and assets.

git clone https://github.com/johnlui/PPHC.git
cd PPHC
ls

If you prefer the rendered version, the README gives the reading address directly, and states that the PDF download link is in the top right corner of the site. Nothing in the README documents a local Markdown-to-PDF pipeline, so downloading from the site is the path the project itself describes.

The README also points at the print edition, published by 电子工业出版社, with a list of material added for print: AI architecture, high availability, distributed theory, an engineering methodology, deeper low-level principles, and a full-stack practical project. The print edition is sold through JD at ¥67.6 and directly by the author at ¥55 for a signed, shrink-wrapped copy, per the README's purchase table. The v1.0.0 release note on 2026-03-06 marks the print publication.

Where PPHC Stops Being the Right Tool

The most obvious limitation is language. The book is written in Chinese, the README is in Chinese, and there is no indication of an official English translation in the repository. An engineer who cannot read Chinese gets nothing from the text files, and the hosted site does not change that.

Second, there is no code to run. The repository holds prose and images. If you want a reference implementation of an epoll server, a Paxos library, or a DPDK data path, this is the wrong repository, and the README does not claim otherwise. The print edition adds a full-stack practical project, but that is a book chapter, not a runnable artifact you can pull.

Third, the licence is restrictive in a specific way. CC BY-NC-ND 4.0 allows sharing with attribution but forbids commercial use and forbids derivative works. That rules out translating the book into another language and publishing the translation, and it rules out incorporating chapters into paid training material. For a free book this is a coherent choice, but it means PPHC cannot function as a base text that a team rewrites for its own internal handbook.

Fourth, the breadth is a trade-off. A single volume that moves from switches to CXL to Paxos will treat each topic at essay depth rather than at reference depth. Readers who need to configure Keepalived or size a DPDK ring should expect to go elsewhere for the operational detail, because the README's own framing is a philosophical principle applied across many layers, not a manual for any one of them.

PPHC Against a Conventional Systems Textbook

The natural comparison is a standard distributed-systems or operating-systems textbook, and the difference is in the organising principle. A conventional textbook is organised by abstraction layer: processes, memory, networks, consensus, each in its own chapter with formal definitions and exercises. PPHC is organised by a single repeated move, find the single point and split it, and it applies that move across layers that a textbook would keep separate. The result is that Nginx, LVS, Paxos and CXL appear in one continuous argument rather than in four disconnected chapters.

The cost of that approach is rigour. A textbook gives you the Paxos paper's assumptions, the failure model, and the proof sketch. PPHC, based on the README's description, gives you the place Paxos occupies in a splitting strategy. If you are implementing consensus, you want the formal source. If you are deciding where a single point sits in your own architecture, the PPHC framing is the more directly useful one.

A second comparison point is the print edition of the same book. The README lists what print adds, and the additions are not cosmetic: AI architecture, high availability, distributed theory, engineering methodology, deeper low-level material, and a full-stack project. The open-source edition is the 12-chapter, 83-article version. Choosing between them is choosing between free and current, since the print edition is the one that received the 2026-03-06 v1.0.0 release.

Licence Terms and What They Mean for Reuse

The README states the copyright belongs to Lü Wenhan and that the book is open sourced under CC BY-NC-ND 4.0, for free reading by GitHub users. The repository metadata reports the licence as NOASSERTION, which means GitHub's detector could not classify the LICENSE file automatically; the README's explicit CC BY-NC-ND 4.0 statement is the clearer signal, and the LICENSE file at the repository root is the authoritative text to read.

The three components matter separately. Attribution means you must credit the author when you share. NonCommercial means you cannot use the material commercially, which includes paid courses and paid internal training built on the chapters. NoDerivatives means you cannot distribute a modified version, which is the clause that blocks an unofficial translation. Reading it, linking to it, and quoting it with attribution are the uses the licence clearly supports.

This is not legal advice, and the practical question for a team is narrow: does anything you plan to do involve redistribution or modification? If the answer is no, the licence is not an obstacle. If the answer is yes, read the licence text at creativecommons.org before you act, because the ND clause is the one people miss.

Maintenance, Upgrades and the Cost of Following the Book

The repository is not archived. The last push was on 2026-06-02, which is recent enough that the text is being touched, though the README gives no changelog and no statement about how updates are scheduled. The only release in the list is v1.0.0 on 2026-03-06, tagged with the note that the print book was published successfully. There is no versioning scheme for the text itself, so if you clone today and clone again in six months, you have no documented way to tell what changed beyond a git log.

Upgrade cost is therefore low in the mechanical sense and uncertain in the editorial sense. Pulling new commits costs nothing. Knowing whether a chapter you cited has been rewritten costs you a diff. For a book that readers cite in architecture discussions, that absence of a changelog is a real gap, and it is the kind of thing a documentation-heavy project usually adds once outside contributions start.

The other cost is the print edition. The README positions print as the more complete version, with AI architecture, high availability, distributed theory, engineering methodology, deeper low-level material and a full-stack project added. A reader who starts with the free edition and later wants those topics buys the book; there is no documented upgrade path from the repository to that added material.

Editorial conclusion

Adopt PPHC if you are a Chinese-reading backend or infrastructure engineer who wants a structured tour from Apache and epoll up to CXL and Paxos, and who wants a free copy before deciding on the print edition. Skip it if you need English text, executable sample code, or permissive licensing for reuse inside a product. Verify first that the five part directories in the repository match the twelve-chapter, 83-article structure the README describes, since the README gives no chapter-to-directory mapping, and confirm the CC BY-NC-ND terms against your intended use before you copy any part of the text.

Frequently asked questions

Official sources

1. Issues
2. johnlui/PPHC on GitHub
3. Project website
4. README
5. Releases