Ensure we are not vulnerable to BREACH attack

My understanding is that we need all of the following conditions to be true for the attack to be an issue:

1. Response body contains sensitive information an attacker may want to guess
2. Response body contains a user given string (such as a search param)
3. HTTP compression is on

Since Funkwhale is an SPA, most of the UI is made of static content. Item 1 is not checked for this resources and we should be able to keep HTTP compression on.

Now, Funkwhale API do contain sensitive data. However, I cannot think of any endpoint that includes sensitive data (1) and user input (such as a search param) in response body. I have to admit this could change in the future though.

Django's admin (/api/admin) is a much more traditionnal web app and may be vulnerable though, since it contains lots of sensitive data (although CSRF were secured with https://code.djangoproject.com/ticket/20869), and possible user input (via search params).

Suggested mitigation:

• Disable HTTP compression for json content
• Disable HTTP compression for admin urls

If anyone wants to give feedback advice on this, you're welcome :)

Links:

• http://breachattack.com/
• https://security.stackexchange.com/questions/172581/to-avoid-breach-can-we-use-gzip-on-non-token-responses
• https://code.djangoproject.com/ticket/20869
• https://github.com/YunoHost/yunohost/pull/356