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- remove the outdated section header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
- 0105*: merge into a single block
- 1220: make values more readable with spaces, like 2701 (no need for value 2), add default, update advise (get a new AV, SHA1 is dead baby)
- 2619: remove fluff
- we already disable webgl, that's enough
- the other two prefs are not going to provide much protection if a user decides they want webgl
- "disable-fail-if-major-performance-caveat" only applies to ESR78 and will removed in the future
- one (or two) less pref(2) for users to troubleshoot/flip
- remove 2720
- this is a very old pref, been inactive since at least our first github release: v51
- disabling the API is not how you control client side state: you do that by blocking cookies which also controls other state such as IDB etc
- 2700 section header
- history/downloads is redundant
- Offline Website Data info -> relevant item number with Active Logins info
- ^ technically it still includes appCache for ESR78 users, but that will be moot in less than three months
- tidy RFP
- update to FF91 userAgent spoofing: there is no Android ESR so we don't need to mention "Android 9"
- we don't need to say if the API is enabled for mediaDevices
- the service implies a check is done first, I'm more concerned with the actual updating: not that updates are bad, it's about controlling when (if ever e.g. my test suite)
- since 0301 has to be done manually in Windows, 0302 is a good fallback **IF** the background service is applicable (read the link)
- clean up the numbering
- "enforce" is for when we set the default value
- use [WARNING] for inactive (they're inactive for a reason and people really do not need to turn them on) but less scary [NOTE] for active (tweak away at your own risk)
- seems neater, easier and less scary for users setting up the first time: i.e they only need to initially look at active items
- FYI: I was going to add something to LSNG (2760) that it is required for Fission, but will wait, and it struck me that 2680 was the only active item with a warning: seems inconsistent
- 2684: security delay .. make enforce mean enforce (default) ... not worth occasionally saving .3 seconds
- for now it's one less item in differences/flips
- might make this inactive in 91+, and add a warning
- it has been a very long time since we added this due to bad advise/references on the internet on how to speed up Firefox
- they all have on/off switches
- dxr no longer exists: update URL
- don't recommend users delete files
- saves two lines
- they poses zero threat (they have prefs)
- deleting them can causes unwanted console errors/noise
It literally cannot hurt [1], and makes it easier for users to use custom mode with TCP/dFPI. Turning on socialtracking helps gain parity with strict mode
[1] gorhill: https://old.reddit.com/r/firefox/comments/l7xetb/network_priority_for_firefoxs_enhanced_tracking/gl9rn9n/
> All extensions and ETP work in parallel, they all inspect network requests and all make the decision to block or not, hence if they all decide to block, they will all report that they block something. ETP is a bit different than normal extension in that it will give precedence to an extension trying to redirect to a local resource, this ensures ETP works harmoniously with normal extensions.
>
> Once something is not blocked, it then goes through a DNS query, and the browser waits for the response.
>
> I will add examples of how ETP + multiple blocker extensions work together when dealing with a network request; let's say "A" and "B" are two different blockers:
>
> - ETP=block, A=allow, B=allow: result=block
> - ETP=allow, A=block, B=allow: result=block
> - ETP=allow, A=allow, B=redirect: result=redirect
> - ETP=allow, A=block, B=redirect: result=block
> - ETP=block, A=allow, B=redirect: result=redirect
>
> So as you can see, ETP is a bit different than a normal extension in that it won't prevent redirection from happening if ever a network request is redirected by one of the normal extension.
It is simpler to leave the PointerEvent pref where it is, until ESR78 is EOL
- FF87+ users who use RFP Alts simply add a dead pref, no harm
- This way ESR78 users don't have to worry about extra char flipping: it's the same as before: 1 flip for ESR, 1 flip for RFP Alts
only stable is false, at the time of writing. but enforcing this for all channels is good, so no-one ends up wasting mozilla resources reporting a compat problem when they've got 200 odd prefs flipped
- don't differentiate between channels
- both can be made inactive
- webcompat requires user action: and I don't see this as a bad thing to have in non-stable
- unsubmitted crashReports on Nightly is probably already covered by killing the URL, so no big deal
- 0000: remove old XUL info, dropped in FF73+
- 0201: save 3 chars
- 0350: add default status for unsubmittedCheck
- 0351: change to enforce: has been default false going back to at least FF60, including current Beta/Dev/Nightly
- along with 0602 `network.dns.disablePrefetchFromHTTPS` and 0603 `network.predictor.enable-prefetch`, I considered making them inactive, but decided it was good to leave them active for non-stable users just in case they get flipped
- 0515: add default status
- 0850c: remove info: out of date: doesn't work lilke that anymore and can't be assed figuring it out what with megabar and urlbar2 changes
- 0871: make inactive: default false since at least FF60
- no need to enforce for non-stable in case it is flipped. It's a pretty minor shoulder-surfer privacy issue and the previews are small. If you're not sure what this pref does. On false you get one tab shown, on true you get as many as can fit across your screen. I squeezed in 15, and after that it became a list
- fixup `***/`
- shave off six lines and almost 400 bytes for you bastards
- This is too minimal to be of any use, breaks too much (e.g. zoom video)
- Tor browser stopped flipping this (I *think*) about 5 years ago: it certainly hasn't been used in ESR60+ based TB builds, I checked
- we already disable webgl, so making this inactive removes yet another pref users need to flip/troubleshoot
- I will leave it in the user js for a few releases so prefsCleaner will pick it up
- It is controlled in both runtime and via user.js by the state of `media.eme.enabled`. Also, who cares about the vis of a ui option
- note, there is no need to add this to the removed scratchpad list
- remove useless `see` word for reference links
- fixup 0701
- "do not play nice" is not measurable
- don't reference to self as a source: people can just search "VPN leak Ipv6" or something
- shrink and remove outdated info from section 0300 header
- combine some bugzillas
- drop some references
- 1647829 for HTTPS-Only mode
- hardware metrics: not going to implicitly encourage users to use this pref or tell them what sizes to use
- update [STATS]
- also remove TLS [STATS].. stats on TLS 1.0 and 1.1 are irrelevant: the default is now TLS 1.2+
- single CRLite reference for all blog articles
- save 588 bytes so all you bastards can theoretically load Firefox just that tiny bit faster
Thorin-Oakenpants