curl/docs/CURLDOWN.md

curldown

A markdown-like syntax for libcurl man pages.

Purpose

A text format for writing libcurl documentation in the shape of man pages.

Make it easier for users to contribute and write documentation. A format that is easier on the eye in its source format.

Make it harder to do syntactical mistakes.

Use a format that allows creating man pages that end up looking exactly like the man pages did when we wrote them in nroff format.

Take advantage of the fact that people these days are accustomed to markdown by using a markdown-like syntax.

This allows us to fix issues in the nroff format easier since now we generate them. For example: escaping minus to prevent them from being turned into Unicode by man.

Generate nroff output that looks (next to) identical to the previous files, so that the look, existing test cases, HTML conversions, existing infrastructure etc remain mostly intact.

Contains meta-data in a structured way to allow better output (for example the see also information) and general awareness of what the file is about.

File extension

Since curldown looks similar to markdown, we use .md extensions on the files.

Conversion

Convert from curldown to nroff with cd2nroff. Generates nroff man pages.

Convert from nroff to curldown with nroff2cd. This is only meant to be used for the initial conversion to curldown and should ideally never be needed again.

Convert, check or clean up an existing curldown to nicer, better, cleaner curldown with cd2cd.

Mass-convert all curldown files to nroff in specified directories with cdall:

cdall [dir1] [dir2] [dir3] ..

Known issues

The cd2nroff tool does not yet handle italics or bold where the start and the end markers are used on separate lines.

The nroff2cd tool generates code style quotes for all .fi sections since the nroff format does not carry a distinction.

Format

Each curldown starts with a header with meta-data:

---
c: Copyright (C) Daniel Stenberg, <daniel@haxx.se>, et al.
SPDX-License-Identifier: curl
Title: CURLOPT_AWS_SIGV4
Section: 3
Source: libcurl
Protocol:
  - HTTP
See-also:
  - CURLOPT_HEADEROPT (3)
  - CURLOPT_HTTPAUTH (3)
TLS-backend:
  - [name]
---

All curldown files must have all the headers present and at least one See-also: entry specified.

If the man page is for section 3 (library related). The Protocol list must contain at least one protocol, which can be * if the option is virtually for everything. If * is used, it must be the only listed protocol. Recognized protocols are either URL schemes (in uppercase), TLS or TCP.

If the Protocol list contains TLS, then there must also be a TLS-backend list, specifying All or a list of what TLS backends that work with this option. The available TLS backends are:

• BearSSL
• GnuTLS
• mbedTLS
• OpenSSL (also covers BoringSSL, libressl, quictls, AWS-LC and AmiSSL)
• rustls
• Schannel
• Secure Transport
• wolfSSL
• All: all TLS backends

Following the header in the file, is the manual page using markdown-like syntax:

    # NAME
    a page - this is a page descriving something

    # SYNOPSIS
    ~~~c
    #include <curl/curl.h>

    CURLcode curl_easy_setopt(CURL *handle, CURLOPT_AWS_SIGV4, char *param);
    ~~~

Quoted source code should start with ~~~c and end with ~~~ while regular quotes can start with ~~~ or just be indented with 4 spaces.

Headers at top-level # get converted to .SH.

nroff2cd supports the ## next level header which gets converted to .IP.

Write bold words or phrases within ** like:

This is a **bold** word.

Write italics like:

This is *italics*.

Due to how man pages do not support backticks especially formatted, such occurrences in the source are instead just using italics in the generated output:

This `word` appears in italics.

When generating the nroff output, the tooling removes superfluous newlines, meaning they can be used freely in the source file to make the text more readable.

To make sure curldown documents render correctly as markdown, all literal occurrences of < or > need to be escaped by a leading backslash.

symbols

All mentioned curl symbols that have their own man pages, like curl_easy_perform(3) are automatically rendered using italics in the output without having to enclose it with asterisks. This helps ensuring that they get converted to links properly later in the HTML version on the website, as converted with roffit. This makes the curldown text easier to read even when mentioning many curl symbols.

This auto-linking works for patterns matching (lib|)curl[^ ]*(3).