Skip to content
This repository has been archived by the owner on Oct 24, 2020. It is now read-only.

Latest commit

 

History

History
332 lines (204 loc) · 22.1 KB

CONTRIBUTING.md

File metadata and controls

332 lines (204 loc) · 22.1 KB

Welcome!

Welcome, and thank you for your interest in contributing to HTTPS Everywhere! HTTPS Everywhere depends on the open source community for its continued success, so any contribution is appreciated.

One of the things that makes it easy to contribute to HTTPS Everywhere is that you don't have to be a coder to contribute. That's because HTTPS Everywhere's most important component is the list of rules that tell it when it can request a website over HTTPS. These rules are just XML files that contain regular expressions, so if you can write XML and simple regexes, you can help us add rules and increase HTTPS Everywhere's coverage. No coding skills necessary!

If you want to have the greatest impact, however, you can help be a ruleset maintainer. Ruleset maintainers are volunteers who examine rulesets contributed by others and work with them to ensure that these rulesets work properly and are styled correctly before they're merged in. While we currently have a couple of extremely dedicated and extremely proficient ruleset maintainers, the backlog of sites to add to HTTPS Everywhere just keeps growing, and they need help! If this sounds interesting to you, please contact us at https-everywhere-rules-owner [at] eff <dot> org.

If you get stuck we have two publicly-archived mailing lists: the https-everywhere list (https://lists.eff.org/mailman/listinfo/https-everywhere) is for discussing the project as a whole, and the https-everywhere-rulesets list (https://lists.eff.org/mailman/listinfo/https-everywhere-rules) is for discussing the rulesets and their contents, including patches and git pull requests.

You can also find more information on about HTTPS Everywhere on our FAQ page.

Thanks again, and we look forward to your contributions!

HTTPS Everywhere Source Code Layout

There are several main areas of development on HTTPS Everywhere: the rulesets, the core codebase, utilities, and tests.

The rulesets can be found in the rules top-level path and include all the rules for redirecting individual sites to HTTPS. These are written in XML. If you want to get started contributing to HTTPS Everywhere, we recommend starting here.

The core codebase consists of the code that performs the redirects, the UI, logging code, the decentralized SSL observatory code, and ruleset loading, and is written in JavaScript. It is divided into two separate extensions. One is based on the XPCOM API used by Mozilla Firefox (soon to be deprecated, located in the src top-level path), and the other is based on the WebExtensions API (located in chromium), first developed in Google's Chromium browser.

The utilities (utils top-level path) include scripts that build the extension, sanitize and perform normalization on rulesets, simplify rules, and help label GitHub issues. Historically, these utilities have been written in Python. Many of the newer utilities are written in JavaScript, and are meant to be run in node. Some of the wrappers for these utilities are in shell scripts.

Tests are performed in headless browsers and located in the test top-level path. These are written in Python, and some of the wrappers for these tests are in shell scripts.

Submitting Changes

To submit changes, open a pull request from our GitHub repository.

HTTPS Everywhere is maintained by a limited set of staff and volunteers. Please be mindful that we may take a while before we're able to review your contributions.


Contributing Rulesets

General Info

Thanks for your interest in contributing to the HTTPS Everywhere rulesets! There's just a few things you should know before jumping in. First some terminology, which will help you understand how exactly rulesets are structured and what each one contains:

  • ruleset: a scope in which rules, targets, and tests are contained. rulesets are usually named after the entity which controls the group of targets contained in it. There is one ruleset per XML file within the src/chrome/content/rules directory.
  • target: a Fully Qualified Domain Name which may include a wildcard specified by *. on the left side, which rules are applied to. There may be many targets within any given ruleset.
  • rule: a specific regular expression rewrite that is applied for all matching targets within the same ruleset. There may be many rules within any given ruleset.
  • test: a URL for which a request is made to ensure that the rewrite is working properly. There may be many tests within any given ruleset.
<!--
        An example ruleset. Note that this example doesn't necessarily
        satisfy the style criteria described below - we just have it
        here to show you what the components of a ruleset looks like.
-->
<ruleset name="eff.org">
        <target host="*.eff.org" />

        <rule from="^http:" to="https:" />

        <test url="http://www.eff.org/https-everywhere/" />
</ruleset>

HTTPS Everywhere includes tens of thousands of rulesets. Any one of these sites can change their HTTPS configuration at any time, so keeping HTTPS Everywhere usable is a task that requires constant maintenance. At the same time, HTTPS deployment on the web is becoming more and more widespread, thanks to projects like Let's Encrypt. This is a very good thing, as it means the web is becoming a safer place! However, with each new ruleset that HTTPS Everywhere includes comes with an increase in both download size upon install and memory usage at runtime. Rather than adding new rulesets, we encourage potential contributors to look for broken rulesets and try to fix them first.

Some rulesets have the attribute platform="mixedcontent". These rulesets cause problems in browsers that enable active mixed-content (loading insecure resources in a secure page) blocking. When browsers started enforcing active mixed-content blocking, some HTTPS sites started to break. That's why we introduced this tag - it disables those rulesets for browsers blocking active mixed content. It is likely that many of these sites have fixed this historical problem, so we particularly encourage ruleset contributors to fix these rulesets first:

git grep -i mixedcontent src/chrome/content/rules

New Rulesets

If you want to create new rulesets to submit to us, we expect them to be in the src/chrome/content/rules directory. That directory also contains a useful script, make-trivial-rule, to create a simple ruleset for a specified domain. There is also a script called utils/trivial-validate.py, to check all the pending rulesets for several common errors and oversights. For example, if you wanted to make a ruleset for the example.com domain, you could run:

cd src/chrome/content/rules
bash ./make-trivial-rule example.com

This would create Example.com.xml, which you could then take a look at and edit based on your knowledge of any specific URLs at example.com that do or don't work in HTTPS. Please have a look at our Ruleset Style Guide below, where you can find useful tips about finding more subdomains. Our goal is to have as many subdomains covered as we can find.

Minimum Requirements for a Ruleset PR

There are several volunteers to HTTPS Everywhere who have graciously dedicated their time to look at the ruleset contributions and work with contributors to ensure quality of the pull requests before merging. It is typical for there to be several back-and-fourth communications with these ruleset maintainers before a PR is in a good shape to merge. Please be patient and respectful, the maintainers are donating their time for no benefit other than the satisfaction of making the web more secure. They are under no obligation to merge your request, and may reject it if it is impossible to ensure quality.

In the back-and-fourth process of getting the ruleset in good shape, there may be many commits made. It is this project's convention to squash-and-merge these commits into a single commit before merging into the project. If your commits are cryptographically signed, we may ask you to squash the commits yourself in order to preserve this signature. Otherwise, we may squash them ourselves before merging.

We prefer small, granular changes to the rulesets. Not only are these easier to test and review, this results in cleaner commits.

Testing

A general workflow for testing sites that provide both HTTP and HTTPS follows. Open a version of the browser of your choice without HTTPS Everywhere loaded to the HTTP endpoint, alongside the browser with the latest code and rulesets for HTTPS Everywhere loaded to the HTTPS endpoint (as described in README.md.) Click around and compare the look and functionality of both sites.

If something fails to load or looks strange, you may be able to debug the problem by opening the network tab of your browser debugging tool. Modify the ruleset until you get it in a good state - you'll have to re-run the HTTPS Everywhere-equipped browser upon each change.

Ruleset Style Guide

Motivation

Rules should be written in a way that is consistent, easy for humans to read and debug, reduces the chance of errors, and makes testing easy.

To that end here are some style guidelines for writing or modifying rulesets. They are intended to help and simplify in places where choices are ambiguous, but like all guidelines they can be broken if the circumstances require it.

Indentation & Misc Stylistic Conventions

Use tabs for indentation. For tests and exclusions, place them under the target that they refer to, indented one additional layer. See below for an example.

We provide an .editorconfig file in the top-level path, which you can configure your editor of choice to use. This will enforce proper indentation.

Use double quotes (", not ').

Wildcards in Targets

Left-Wildcards

Avoid using the left-wildcard (<target host='*.example.com' />) unless you intend to rewrite all or nearly all subdomains. If it can be demonstrated that there is comprehensive HTTPS coverage for subdomains, left-wildcards may be appropriate. Many rules today specify a left-wildcard target, but the rewrite rules only rewrite an explicit list of hostnames.

Instead, prefer listing explicit target hosts and a single rewrite from "^http:" to "^https:". This saves you time as a ruleset author because each explicit target host automatically creates an implicit test URL, reducing the need to add your own test URLs. These also make it easier for someone reading the ruleset to figure out which subdomains are covered.

If you know all subdomains of a given domain support HTTPS, go ahead and use a left-wildcard, along with a plain rewrite from "^http:" to "^https:". Make sure to add a bunch of test URLs for the more important subdomains.

Edge-Case: Right-Wildcards

Right-wildcards (<target host='account.google.*' />) are highly discouraged. Only use them in edge-cases where other solutions are unruly.

Example:

Where they must be used, please add a comment to the ruleset explaining why.

Complicated Regex in Rules

Avoid regexes with long strings of subdomains, e.g. <rule from="^http://(foo|bar|baz|bananas).example.com" />. These are hard to read and maintain, and are usually better expressed with a longer list of target hosts, plus a plain rewrite from "^http:" to "^https:".

In general, avoid using open-ended regex in rules. In certain cases, open-ended regex may be the most elegant solution. But carefully consider if there are other options.

Examples:

  • Rulesets with a lot of domains that we can catch with a simple regex that would be tedious and error-prone to list individually, like 360.cn.xml
  • CDNs with an arbitrarily large number of subdomains, like EFForg#7484 (comment) .

Enumerating Subdomains

If you're not sure what subdomains might exist, you can install the Sublist3r tool:

git clone https://github.com/aboul3la/Sublist3r.git
cd Sublist3r
sudo pip install -r requirements.txt # or use virtualenv...

Then you can to enumerate the list of subdomains:

python sublist3r.py -d example.com -e Baidu,Yahoo,Google,Bing,Ask,Netcraft,Virustotal,SSL

Alternatively, you can iteratively use Google queries and enumerate the list of results like such:

  1. site:*.eff.org
  2. site:*.eff.org -site:www.eff.org
  3. site:*.eff.org -site:www.eff.org -site:ssd.eff.org

... and so on.

Target Ordering

In all cases where there is a list of domains, sort them in alphabetical order starting from the top level domain at the right reading left, moving ^ and www to the top of their group. For example:

example.com
www.example.com
a.example.com
www.a.example.com
b.a.example.com
b.example.com
example.net
www.example.net
a.example.net

Rule Ordering

If there are a handful of tricky subdomains, but most subdomains can handle the plain rewrite from "^http:" to "^https:", specify the rules for the tricky subdomains first, and then then plain rule last. Earlier rules will take precedence, and processing stops at the first matching rule. There may be a tiny performance hit for processing exception cases earlier in the ruleset and the common case last, but in most cases the performance issue is trumped by readability.

Non-working hosts

It is useful to list hosts that do not work in the comments of a ruleset. This is a stylistic preference but is not strictly required.

For easy reading, please avoid using UTF characters unless in the rare instances that they are part of the hostname itself.

Example:

<!--
        Invalid certificate:
                8marta.glavbukh.ru
                forum2.glavbukh.ru (incomplete certificate chain)

        Redirect to HTTP:
                8marta2013.glavbukh.ru
                den.glavbukh.ru

        Refused:
                e.glavbukh.ru
                www.e.glavbukh.ru

        Time out:
                psd.glavbukh.ru
                str.glavbukh.ru

-->

In most cases, the absence of a 2XX or 3XX endpoint indicates that a host should not be included in the set of targets and is non-working, except when it is clear that the site functions as intended in the absence of such an endpoint.

Ruleset Names

For simple sites, the ruleset name attribute can be either a site description or the domain itself. For example, the SeattleAquarium.org.xml ruleset could have a name of Seattle Aquarium, SeattleAquarium.org, or seattleaquarium.org.

If a ruleset covers multiple domains, then the ruleset name should reflect the broader organization, project, or concept for what a ruleset is trying to accomplish.

Examples:

Filenames

Filenames should vaguely resemble the name so that someone looking for the file based on the name can find it easily. Filenames that start with a capital letter are preferred. Prefer dashes over underscores in filenames. Dashes are easier to type.

Cross-referencing Rulesets

This sort of comment: For other Migros coverage, see Migros.xml. is definitely appropriate, in both directions.

Regex Conventions

When matching an arbitrary DNS label (a single component of a hostname), prefer ([\w-]+) for a single label (i.e. www), or ([\w.-]+) for multiple labels (i.e. www.beta). Avoid more visually complicated options like ([^/:@\.]+\.)?.

For securecookie tags, if you know that all cookies on the included targets can be secured (which in particular means that the cookies are not used by any of its non-securable subdomains), use the trivial

<securecookie host=".+" name=".+" />

where we prefer .+ over .* and .. They are functionally equivalent, but it's nice to be consistent.

Avoid the negative lookahead operator ?!. This is almost always better expressed using positive rule tags and negative exclusion tags. Some rulesets have exclusion tags that contain negative lookahead operators, which is very confusing.

Prefer capturing groups (www\.)? over non-capturing (?:www\.)?. The non-capturing form adds extra line noise that makes rules harder to read. Generally you can achieve the same effect by choosing a correspondingly higher index for your replacement group to account for the groups you don't care about.

Snapping Redirects

Avoid snapping redirects. For instance, if https://foo.fm serves HTTPS correctly, but redirects to https://foo.com, it's tempting to rewrite foo.fm to foo.com, to save users the latency of the redirect. However, such rulesets are less obviously correct and require more scrutiny. And the redirect can go out of date and cause problems. HTTPS Everywhere rulesets should change requests the minimum amount necessary to ensure a secure connection.

Example: Ruleset before style guidelines are applied

<ruleset name="WHATWG.org">
  <target host='whatwg.org' />
  <target host="*.whatwg.org" />

  <rule from="^http://((?:developers|html-differences|images|resources|\w+\.spec|wiki|www)\.)?whatwg\.org/"
    to="https://$1whatwg.org/" />
</ruleset>

Example: Ruleset after style guidelines are applied, with test URLs

<ruleset name="WHATWG.org">
	<target host="whatwg.org" />
	<target host="www.whatwg.org" />
	<target host="developers.whatwg.org" />
	<target host="html-differences.whatwg.org" />
	<target host="images.whatwg.org" />
	<target host="resources.whatwg.org" />
	<target host="*.spec.whatwg.org" />
		<test url="http://html.spec.whatwg.org/" />
		<test url="http://fetch.spec.whatwg.org/" />
		<test url="http://xhr.spec.whatwg.org/" />
		<test url="http://dom.spec.whatwg.org/" />
	<target host="wiki.whatwg.org" />

	<rule from="^http:"
		to="https:" />
</ruleset>

Removal of Rules

Regular Rules

It should be considered a sufficient condition for removal if a contributor can demonstrate that the TLS configuration for either a specific target or a ruleset altogether is unstable and/or breaking, or will be unstable and/or breaking in the near future. It is, of course, preferable that the ruleset be fixed rather than removed.

HSTS Preloaded Rules

In utils we have a tool called hsts-prune which removes targets from rulesets if they are already contained in the HSTS preload list for browsers that we support. To be explicit, the script is an implementation of the following policy:

Let included domain denote either a target, or a parent of a target. Let supported browsers include the ESR, Dev, and Stable releases of Firefox, and the Stable release of Chromium. If included domain is a parent of the target, the included domain must be present in the HSTS preload list for all supported browsers with the relevant flag which denotes inclusion of subdomains set to true. If included domain is the target itself, it must be included the HSTS preload list for all supported browsers. Additionally, if the http endpoint of the target exists, it must issue a 3XX redirect to the https endpoint for that target. Additionally, the https endpoint for the target must deliver a Strict-Transport-Security header with the following directives present:

  • max-age >= 10886400
  • includeSubDomains
  • preload

If all the above conditions are met, a contributor may remove the target from the HTTPS Everywhere rulesets. If all targets are removed for a ruleset, the contributor is advised to remove the ruleset file itself. The ruleset rule and test tags may need to be modified in order to pass the ruleset coverage test.

Every new pull request automatically has the hsts-prune utility applied to it as part of the continual integration process. If a new PR introduces a target which is preloaded, it will fail the CI test suite. See:

  • .travis.yml
  • test/travis.sh

Contributing Code

In addition to ruleset contributions, we also encourage code contributions to HTTPS Everywhere. There are a few considerations to keep in mind when contributing code.

Officially supported browsers:

  • Firefox Stable
  • Firefox ESR
  • Chromium Stable

We also informally support Opera browser, but do not have tooling around testing Opera. Firefox ESR is supported because this is what the Tor Browser, which includes HTTPS Everywhere, is built upon. For the test commands, refer to README.md.

Several of our utilities and our full test suite is written in Python. Eventually we would like the whole codebase to be standardized as JavaScript. If you are so inclined, it would be helpful to rewrite the tooling and tests into JavaScript while maintaining the functionality.


Contributing Translations

HTTPS Everywhere translations are handled through Transifex. The easiest way to help with translations is to create a Transifex account if you don't already have one. Then log into your account and click "Explore", then search for "Tor Project", and click on The Tor Project. Then choose the language you plan to translate into, click on the name of that language, and then click "Join team" and "Go" to accept joining the translation team for your language.

Then, in the Tor Project resources list, find and click the link for the file

HTTPS Everywhere - https-everywhere.dtd

and choose "Translate now" to enter the translation interface.