Vonng/ddia: reading and rebuilding the Simplified Chinese DDIA translation

The repository holds a translation, not an edition of the book itself

Vonng/ddia is the Simplified Chinese translation of Designing Data-Intensive Applications, second edition, authored by Martin Kleppmann and Chris Riccomini, and it is published as a website at ddia.vonng.com rather than as a packaged release. The repository has no GitHub releases at all. What it contains is the text, organized into fourteen chapters across three parts, plus a glossary and a colophon. Part one, on the foundations of data systems, covers trade-offs in data system architecture, defining nonfunctional requirements, data models and query languages, storage and retrieval, and encoding and evolution. Part two covers replication, partitioning, transactions, the unpleasantness in distributed systems, and consistency and consensus. Part three covers batch processing, stream processing, the philosophy of stream processing systems, and doing the right thing.

Building it needs Hugo, and only the dev target reads a local theme checkout

The site is a Hugo project that leans on an external theme module. go.mod declares module github.com/Vonng/ddia, requires Go 1.27.3, and pins github.com/pgsty/oink v1.4.0 as an indirect dependency, which is the theme. The default make target is dev, and that target alone is special: it sets a module replacement so Hugo reads the theme from a local OINK checkout at $HOME/pgsty/oink instead of the version go.mod pins, which is how a theme change becomes visible without cutting a tag. It then runs:

hugo server --renderToMemory -DFE

where the flags show drafts, future, and expired pages and keep output in memory instead of writing to public/. Every other target keeps the pinned module. The consequence for readers building this locally is that a plain clone can run build, serve, and check, but the dev target needs that separate OINK checkout on disk first.

The serve target disables fast render, and check panics on any warning

The Makefile splits preview from verification, and both ends are deliberately strict. For a production-shaped preview there is serve, which runs `hugo serve --environment production --minify --disableFastRender --disableLiveReload`, so what you browse is minified and rebuilt from scratch rather than patched incrementally as you type. For verification there is check, which first runs `GOWORK=off go mod verify` to confirm the module graph with no workspace in play, then invokes hugo with --cleanDestinationDir, --printPathWarnings and --printI18nWarnings, and finally runs bin/check-rendered-links.py --all. The consequence for anyone editing content is that path and translation warnings are printed and rendered links are checked on every check run. A check-local variant applies the same warning settings while substituting the local theme replacement, so theme edits get the same scrutiny as committed ones.

EPUB output and figure layout each carry a check target

Generated artifacts get verification targets instead of manual review. The epub target runs `bin/epub`, and epub-check depends on it before running `bin/check-epub.py`, so the electronic book is rebuilt and then inspected rather than inherited from an earlier build. Figures work the same way with a write and a check split: figures runs `bin/figure-layout.py --write`, and figures-check runs the same script in check mode, as `bin/figure-layout.py --check`. The consequence for contributors is that layout is owned by scripts under bin/, so the correct way to change a figure position or an EPUB manifest is to edit the script and regenerate, not to patch the generated file by hand and hope the next build agrees. Note also that the repository registers Python as its primary language, which comes from those bin/ scripts rather than from the translated prose.

The Traditional Chinese edition is generated by a pinned OpenCC run

The Traditional Chinese version is not translated a second time by hand. It is produced by a conversion script invoked through uv against a pinned OpenCC implementation, so the conversion toolchain is reproducible: `uv run --with opencc-python-reimplemented==0.1.7 -- python bin/zh-tw.py --edition v2`. Pinning that exact version matters, because an unpinned dependency would let the same input produce different output as the converter evolves underneath you. The repository credits @afunTW with both the Traditional Chinese version and the conversion script. The consequence for readers and contributors is that the zh-tw text is a build product derived from the Simplified source. Direct edits to converted files are exposed to being overwritten the next time the target runs, so changes belong in the source or in the script.

The legal notice forbids redistributing the translation or using it commercially

The project's own legal notice is the strictest limit on what this repository is for. The translator states that the work was done purely for study purposes and personal interest with no economic benefit sought, that attribution is retained in this version while other rights remain with the original author and publisher, and that the translation is provided for study and research reference only. It must not be publicly distributed or used for commercial purposes, and anyone able to read English is directed to buy the legitimate edition, with the English original offered through O'Reilly. The notice also records that the original author had mentioned a separate Simplified Chinese translation plan intended to finish at the end of 2018. The consequence for readers is concrete: this text cannot be republished, mirrored as a corpus, or bundled into a paid product.

There are no tagged releases, so the text can change under you

Because the repository has no GitHub releases, there is no version to pin. Everything lives on the default branch, and the project describes itself as under continuous revision: a note in the file states that all fourteen chapters of the second edition are online and are still being proofread, and invites readers to read and offer comments. The last push to the repository was November 2, 2026, so the text is still changing. The proofing is real work rather than a formality, with full-text proofreading credited to @yingang, chapter corrections from contributors including @seagullbird, @nevertiree, @MuAlex, @jiajiadebug, and @cg-zhou, and the Traditional Chinese conversion from @afunTW, all inside a contributor list of 212 names. The consequence for anyone citing the translation is that a passage can be revised without notice and there is no immutable snapshot to point at.

Editorial conclusion

Use Vonng/ddia if you read Chinese and want the DDIA second edition with stable figure numbering, cross references, sequential reading, whole-book printing, Markdown and llms.txt output, and EPUB export, whether you read it at ddia.vonng.com or build the site yourself with Hugo and the OINK theme. Do not treat it as a redistributable corpus: the legal notice confines the translation to study and research, excludes commercial use, and points English readers to the O'Reilly edition, so verify the terms before quoting it in published work. Also plan for moving text, because there are no tagged releases and the chapters are under continuous revision.

Frequently asked questions

Official sources

1. Issues
2. License: CC-BY-4.0
3. Project website
4. README
5. Vonng/ddia on GitHub