More minor tweaks to come. This isn't final
- 0102: ambiguous that the clearing was related to PB mode
- 0900s:
- get rid of 0901, it has no pref, stick link in header
- 0905: values on multi-lines use spaces = more readable
- 1000s:
- rename as disk avoidance and remove sub-section headers
- remove the outdated section header
- 4001: it will never be perfected, it's doing it's job
- 5500s: optional hardening
- legit security measures, but commonality in caveats, so I made them a separate section
- this flips graphite, asm.js and wasm from active to inactive: these are overkill: exhibit A: hundreds of millions of Firefox users
- e.g. graphite and wasm are enabled on Tor Browser
- new CVE keyword links
- 7000s: don't bother - two more items added
- 5000s: optional opsec and cleanout 0800s header
- re-number
- 0900s, 1000s, 1400s, 2400s
PS: I need a new parrot: "9000 syntax error: I ran out of parrots"
Yes it's pretty much useless. Yes it's fingerprintable, and what that entropy is, who knows. Since it's sent regardless with ETP, which we enable in all windows, then who cares. And if you don't use ETP in all windows, then I don't care either - just saying
probably more professional to keep it at the end since it isn't strictly project related. It also opens up space for `DON'T TOUCH` and `OPTIONAL OPSEC`
- merged 3DES cipher to bottom: it is still the same order of [1]
- 3DES pref will be deprecated: pref name changes, and the cipher slated to be unavailable unless you downgrade to < TLS1.2 - see https://bugzilla.mozilla.org/show_bug.cgi?id=1724072
- FYI: we reset TLS downgrades to session only by resetting the pref currently in 1203
- "Minimal/non-existent threat of downgrade attacks"
- FYI: these old ciphers are about 1-2% of traffic (from memory) - but that's still significant breakage
- So the only reason to do this would be to harden against downgrade attacks (and inadvertently use weak sites = breakage): but that doesn't fit most user's threat model: and is probably never going to happen for them. Not sure if I can word that much better and just as succinct
- inactive in user.js since
- v55: gfx.direct2d.disabled
- v67: layers.acceleration.disabled
- the way to counter hardware fingerprinting is within each API that may expose it
- this may have made some sense way back in the day, when there were less options/protections, but not any more
- [are we web render yet](https://arewewebrenderyet.com/) - yes, 100% - there is no need to cripple your browser's perf
- inactive since we added it in v63
- this is not how you defeat fingerprinting (unless done in an enforced set)
- for the record: not even tor browser disable this
- fingerprinting this is not cheap in gecko (for now)
- from [2]
- decoding/encoding capabilities: "it is expected that the entropy ... isn’t going to be significant"
- HDR detection: "... has the potential to add significant entropy .. however .. but ... thus minimizing effective entropy" - it is what it is
- note that RFP has some mitigations in FF82+ 1461454
- just to be clear, this section is not supported: not interested in references or explanations or FF version numbers or default info etc
- "do more harm than good" - ambiguous, not interested in explaining why exactly: but FYI
- some leak
- most break shit
- almost all are easily fingerprinted and the combo of them would make you really stand out
- removed the duplicate `ui.prefersReducedMotion` - this should move to personal as well
- moved `ui.systemUsesDarkTheme` to personal
8000s (was 4600s)
- move below personal, so user-relevant part is shorter
- swap out font vis with document fonts + font whitelist
- font vis still has usability/visual purposes: it just won't really help much with fingerprinting
- ESR78 users (who can't use font vis), sorry, but we made doc fonts inactive for a while now, and now recommend you don't use it anyway
- dead weight since 2017-06-13 when ESR45 reached EOL .. good riddance
- if someone does use it, it's not going to do any harm, so no need to carry it for prefsCleaner
- geo -> warning
- merge container prefs
- remove redundant "see"s
- remove corresponding 4600's item number in RFP mitigations
- it's pretty clear by the preference names in 4600
- could be misconstrued that the 4600 pref is the same result
- RFP's language prompt only checks for en*, not en-US (so en-GB, en-CA etc do not get prompted)
- https://searchfox.org/mozilla-central/source/toolkit/components/resistfingerprinting/RFPHelper.jsm#196
