Skip to main content

Xml2rfc generates RFCs and IETF drafts from document source in XML according to the IETF xml2rfc v2 and v3 vocabularies.

Project description


The IETF uses a specific format for the standards and other documents it publishes as RFCs, and for the draft documents which are produced when developing documents for publications. There exists a number of different tools to facilitate the formatting of drafts and RFCs according to the existing rules, and this tool, xml2rfc, is one of them. It takes as input an xml file which contains the text and meta-information about author names etc., and transforms it into suitably formatted output. The input xml file should follow the grammars in RFC7749 (for v2 documents) or RFC7991 (for v3 documents). Note that the grammar for v3 is still being refined, and changes will eventually be captured in the bis draft for 7991. Changes not yet captured can be seen in the xml2rfc source v3.rng.

xml2rfc provides a variety of output formats. See the command line help for a full list of formats. It also provides conversion from v2 to v3, and can run the preptool on its input.


Installation of the python package is done as usual with ‘pip install xml2rfc’, using appropriate switches and/or sudo.

Installation of support libraries for the PDF-formatter

In order to generate PDFs, xml2rfc uses the WeasyPrint module, which depends on external libaries that must be installed as native packages on your platform, separately from the xml2rfc install.

First, install the Cairo, Pango, and GDK-PixBuf library files on your system. See installation instructions on the WeasyPrint Docs:

(Python 3 is not needed if your system Python is 2.7, though).

(On some OS X systems with System Integrity Protection active, you may need to create a symlink from your home directory to the library installation directory (often /opt/local/lib):

ln -s /opt/local/lib ~/lib

in order for weasyprint to find the installed cairo and pango libraries. Whether this is needed or not depends on whether you used macports or homebrew to install cairo and pango, and the homebrew / macport version.)

Next, install the pycairo and weasyprint python modules using pip. Depending on your system, you may need to use ‘sudo’ or install in user-specific directories, using the –user switch. On OS X in particular, you may also need to install a newer version of setuptools using –user before weasyprint can be installed. If you install with the –user switch, you may need to also set PYTHONPATH, e.g.,


for Python 2.7.

The basic pip commands (modify as needed according to the text above) are:

pip install ‘pycairo>=1.18’ ‘weasyprint<=0.42.3’

With these installed and available to xml2rfc, the –pdf switch will be enabled.

For PDF output, you also need to install the Noto font set. Download the full set from, and install as appropriate for your platform.


xml2rfc accepts a single XML document as input and outputs to one or more conversion formats.

Basic Usage: xml2rfc SOURCE [options] FORMATS...

Run xml2rfc --help for a full listing of command-line options.


Version 3.1.0 (13 Sep 2020)

  • Fixed an issue with an extra comma rendered in HTML for empty date elements. Fixes issue #542.

  • Added escaping of quotes in page footer and header components when generating the @page CSS for PDF output, in order to avoid issues when building quoted CSS content strings from the components. Fixes issue #544.

  • Added a CSS workaround for an unexpected HTML rendering behaviour in some browsers, where a <dt> following an empty <dd> would be indented. Fixes issue #545.

  • Improved error messages and the handling of artwork with no “type” attribute value when “binary-art” would have been correct. Fixes issue #535.

  • Reverted the default value for the –legacy-date-format to false, on request from the RPC.

  • Fixed table cell padding to not add left paddding for left-aligned columns if not all lines in the column can be padded in the same manner, and similarly for right-aligned columns. Fixes issue #543.

  • Tweaked utils.build_dataurl() to give consistent results across Python versions (tilde was added to the default safe characters in Python 3.7, so the default percentage-escape results of binary content differed if it contained a 0x7e (tilde) character.

  • Fixed a bug in finding the length of the longest word in a table cell, used to determine minimum cell width for text output.

  • Added a new test case with a tightly constrained table to excercise the issue in #543.

  • Added <title> to the list of elements that can have Latin script content without needing the ‘ascii’ attribute set. Fixes issue #525.

  • Refined the rendering of <xref> with format=’title’ for the text formatter.

  • Added docfile and manpage information about elements that are permitted to have Latin script content without an ascii attribute equivalent being set.

  • Expanded on the docfile description of elements that can have Latin script content without ASCII fallback.

  • Added a test case related to issue #525.

  • Changed the HTML rendering of <xref> within <name> to use square brackets. Fixes issue #498.

Version 3.0.0 (02 Sep 2020)

Transition to using the new schema v3 output formatters by default

This release provides the functionality that the 2.47.0 release had (with some enhancements), but is backwards incompatible because the default settings for some switches has changed. The –legacy switch must now be set explicitly in order to use the old output formatters. By default, XML input files with schema v2 content will be converted to v3 on the fly and the output formatting of the converted XML will be done with the new schema v3 formatters. With this release, support of Python 2.7, which is past end-of-life, will no longer be part of the test suite.

There are also a number of other changes. From the commit log:

  • Replaced the use of the deprecated optparse module with the newer argparse python module.

  • Removed testing with Python 2.7, and added Python 3.8

  • Updated the bin/mkrelease script to generate documentation HTML and text for the release, place it on, and mention the documentation URL in the release notes.

  • Updated the major revision to 3, given that we no longer support Py27 and have switched default output formatters.

  • Changed bin/mkrelease to install using pip3.6 on the tools servers (the default pip might be for Py2.7).

  • Added an ‘indent’ attribute for <t>, in order to support indented paragraphs without the one-item unordered list workaround, as approved by the schema change board. Added default values for the ‘indent’ attributes for <dl>, <ul>, <ol>, <t>. For the <t> element, the ‘indent’ attribute indicates any extra amount of indentation to be used when rendering the paragraph of text. The indentation amount is interpreted as characters when rendering plain-text documents, and en-space units when rendering in formats that have richer typographic support such as HTML or PDF. One en-space is assumed to be the length of 0.5 em-space in CSS units. Only non-negative integer amounts of indentation are supported.

  • Improved an error message about bad attribute values to show the line of XML source on which the error was found.

  • Added information about command-line switches that have negations (–no-foo… versions) to the context handed to the documentation template.

  • Changed the default of some switches for the 3.0.0 release: –v3 => true; –legacy-date => true; –external-js => false.

  • Improved the documentation file output for switch default values and for options with negation switches.

  • Updated the Makefile to use the appropriate 3.x release series switches.

  • Updated the requirements for a number of python modules.

  • Fixed an issue where hrefs without matching ids could be generated by the HTML renderer from empty <name> elements. This also fixed an issue with missing figure and table captions in some unusual cases.

  • Added support for multi-level ordered lists through a ‘%p’ (for parent) code for use in the <ol> ‘type’ attribute. Fixes issue #465.

  • Added more documentation for the –version switch

  • Updated the schema and tests to permit <blockquote> within <aside>. Fixes issue #524.

  • Added a list of available postal elements for a country to the warning for unused postal address parts.

  • Added a length limitation for the running header title in paginated text documents, to avoid overwriting other parts of the running header.

  • Changed the schema to permit nested <sub> and <sup>, as approved by the v3 schema change board.

  • Added support for outdent handling to propagate upwards to parent elements if the full needed outdent amount could not be done in the local context, in order to be able to apply artwork outdenting to <artwork> elements which aren’t situated immediately under <section>.

  • Changed many instances of reference source indications (xml:base) from “” to just “”. Removed the massaging of reference XML to place seriesInfo elements in the backwards-incomplatible location inside reference/front. Changed the –add-xinclude flag to use as the location of draft reference entries.

  • Added a couple of entries to the test suite reference cache.

  • Improved the handling of missing day information for <date> to make sure we don’t pick days outside the acceptable range for the given month and also pick a reasonable value based on whether the year and month is in the past, present or future.

  • Improved an error message for a case of disallowed XML text content. Tweaked the ‘block_tags’ list.

  • Changed the manpage template to not use comma before ‘and’ when rendering a list of 2 elements.

  • Changed the schema to permit <aside> within <dl> on request from the RPC, with schema change board approval. Updated renderers, CSS and tests accordingly.

  • Tweaked the CSS for block elements that are direct first children of <dd> to render the same way in HTML as in text (i.e., vertically distinct, not on the same line as <dt>).

Version 2.47.0 (17 Jul 2020)

CSS fixes, Built-in documentation, manpage mode, and more

The major feature in this release is the addition of built-in documentation generated from:

  • the actual XML schema distributed with the tool

  • the differences between the current schema and the RFC7991 schema

  • the code’s settings for which elements and attributes are deprecated

  • text snippets describing the schema parts and how the code handles them.

There are new –doc/–docfile and –man/–manpage switches; the first will generate documentation in the form of a v3 XML document that can then itself be processed to generate the various supported formats. The second, –man, will generate the documentation XML internally and then process it to text output which is shown with a pager, like ‘man’.

From the commit log:

  • Corrected the CSS line height of compact lists (it should not be different than for non-compact lists; compact should only affect spacing between items, not line height). Also corrected the CSS top margin for nested lists; extra top margin is desired for a top-level list, but not when nesting them, due to the resulting inconsistency in apparent line height variations.

  • Changed <section> within <toc> from oneOrMore to zeroOrMore, in order to make it possible to honour the tocInclude setting, and reordered some true/false entries for consistency, and changed the line breaking of some lines in the RNG compact representation to fit on 72-character lines.

  • Added .rng and .rnc files with the RFC7991 schema, in order to be able to automatically determine which elements and attributes are new in the schema since 7991.

  • Added a new writer, ‘doc’, and template and text snippet files for autogenerated documentation.

  • Updated the requirements file with some new module requirements.

  • Refactured the code a bit to make it more straightforward to generate text without writing it out to file.

  • Moved the list of deprecated attributes to writers/, and did some slight refactoring for consistent naming of some class variables and avoidance of duplicate parsing of the schema file.

  • Did some minor code cleanup and dead code removal, and corrected the header generation for non-IETF documents (using <rfc ipr=’’>).

  • Fixed an issue where XML parser errors could be reported for ‘<string>’ instead of the actual input file name.

  • Added new options –docfile/–doc and –manpage/–man, used to trigger generation and display of the built-in documentation. Reorganised the option grouping. Updated some option help strings. Made it possible to propagate all command-line option information to the documentation template.

  • Corrected the default templates path. This is related to change [3723].

  • Reverted a change from [3722] in the v2v3 converter.

  • Added a custom Jinja2 filter ‘capfirst()’, for use in the documentation template.

  • Tweaked the documentation template: Some changed wording, support for sub-items not wrapped in <t>, corrected capitalisation using the ‘capfirst’ filter.

  • Updated hastext() and iscomment() to do the right thing if given content with embedded xml processing instructions.

  • Tweaked the handling of default values for –date, so as to give better documentation of the option, and also tweaked the help text for –table-borders.

  • Added a class utility method to get any current PI related to a given setting, and fixed another case of template path default value, related to [3722].

  • Added PI support for text table borders setting, and improved the text table output for transitions between <th> and <td> rows> for ‘light’ and ‘minimal’ borders.

  • Added makefile support for testing of the –manpage and –docfile switches. Added silencing of one unavoidable warning in the test of –unprep.

Project details

Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

xml2rfc-3.1.0.tar.gz (4.5 MB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page