TL;DR: check out my GNU docs builds at

They uniquely have:

Send me an email if you want some page that's not in there, I'll see what I can do now that I have the tools built.

I have recently adopted the practice of locally building and serving HTML documentation for projects I use and work on. I have concluded it's a really good idea:

To achieve this I've set up nginx as a user unit and have stuffed some docs into that web root. Serving docs over http provides a few notable advantages over the file:// protocol: does not lose zoom across pages (accessibility problem!) and can read index.html files when present.

Rust docs

I do the usual routine with cargo doc in a project, which will generate docs with links between libraries and the exact right versions of dependencies. However, there are some ways I've improved my experience:

GNU docs

Preface: GNU is politically useless, they refuse to kick out their misogynist-in-chief despite him not doing anything good or useful for years, and their software is not even very good. I absolutely do not endorse them. However, their software is unfortunately very popular and available everywhere. Also, for instance, LLVM copies the invocation and other details of the GNU equivalents to their tools and does not produce separate documentation, in general. Irrespective of reason, I very often end up needing to read GNU documentation.

The existing ways of reading GNU docs are all terrible. info(1) is hot garbage, with key bindings that nobody will remember. If you choose to look at the HTML docs published by the GNU project, they're in about 50 HTML files each of which is less than a screenful, making them absolutely impossible to search except with like, Google, which also could do way better but doesn't.

It's actually possible to build single page docs, you do something like the below snippet:

$ git clone git://
$ cd binutils-gdb
$ mkdir build
$ cd build
$ ../configure
$ make html MAKEINFO=makeinfo MAKEINFOFLAGS='--no-split'
$ make pdf
$ find . '(' -name '*.pdf' -or -name '*.html' ')' -exec cp '{}' ~/dev/docs/gnu ';'

However, since GNU has development practices stuck in the 90s and no CI, it is an extremely frequent occurrence that their source repo won't build or some other nonsense, or nobody has tested the texinfo options or similar (all of these have happened to me). The point is, this sucks greatly.

Also, when you do actually build these docs, they are using 90s web design practices that make them unusable on modern machines: there is no significant CSS in use, so the lines are about 400 characters long on my machine. I normally fix this with Stylus rules but I've written a BeautifulSoup script to bake some CSS into the texinfo HTML I am publishing.

As of now they are also available publicly at