* Onions to Tor, I2P to I2P. [My rules](https://bafybeia7yeie5a6arstytabl6bfhacmidcy4yhht3e6dfrrq4gi2gry4pu.{{site.ipfsSubdomainGateway}}/foxyproxy.json)
These can generally be found from `about:flags` on Chromium based browsers, for Vivaldi explicit `vivaldi://flags` is required and it also has `chrome://settings` for the usual Chromium settings.
*`#enable-quic` - enabled
*`#enable-force-dark` - enabled with increased text constract
*`#force-color-profile` - sRGB
*`#trust-tokens` - enabled
### Vendor-prefixed
These likely also exist, but just without the `vendor-` part when searhcing.
*`dom.security.https_only_mode` to `true` to force HTTPS and not need HTTPS Everywhere
* [Breaks IPFS companion subdomain gateway redirect](https://github.com/ipfs-shipyard/ipfs-companion/issues/855), see also [Firefox bug 1220810 Consider hardcoding localhost names to the loopback address](https://bugzilla.mozilla.org/show_bug.cgi?id=1220810#c23)
*`security.OCSP.require` to `true` in order to not allow [OCSP](https://en.wikipedia.org/wiki/OCSP_stapling) soft fail. This may be a bit paranoid, but *only the paranoid survive.*
*`geo.provider.network.url` to `https://location.services.mozilla.com/v1/geolocate?key=%MOZILLA_API_KEY%` in order to send nearby WiFi networks to Mozilla instead of Google. See also [MLS Software](https://wiki.mozilla.org/CloudServices/Location/Software).
*<s>`media.peerconnection.enabled` to `false` in order to disable WebRTC (potential IP leaker, will break VoIP/calls, but those are better outside of Firefox anyway)</s>*Not needed anymore in 2020, WebRTC has improved to not do that.*
*`media.navigator.enabled` to `false` in order to also hide cameras and microphones from websites. *I am not sure if this is still necessary either, but maybe it will remind me that I have focused my VoIP to Chromium?*
*`network.IDN_show_punycode` to `true` in order to see punycode instead of UTF-8 in case of spoofing attempt. However makes reading non-ASCII domains painful.
*`reader.parse-on-load.force-enabled` to `true` in order to allow reader use to be used on ~all websites and devices (regardless of low RAM?)
*`toolkit.telemetry.server` to empty in order to not send telemetry (which may be blocked by filtering DNS providers such as AdGuard or NextDNS resulting high amount of failing queries)
*`network.trr.bootstrapAddress` DNS server to use for resolving the DoH
name, e.g. `149.112.112.112` (Resolver 2 of [Quad9](https://quad9.net))
*`network.trr.mode` depends, 2 to prefer DoH, but fallback to system resolver (or 3 to enforce DoH without fallback). ***If there is system encrypted DNS, just take 5 to at least benefit from the system DNS cache.***
* [DoH is required by Firefox ESNI support](https://bugzilla.mozilla.org/show_bug.cgi?id=1500289) which encrypts SNI which would still leak which
sites you visit. [Another bug about ESNI + Android DoT](https://bugzilla.mozilla.org/show_bug.cgi?id=1542754#c3)
* I have ended up to recommending 2 as otherwise the DoH server going
down stops DNS from working on your Firefox entirely, which may be
more of a problem than unencrypted SNI as not everyone supports it.
* since then I have decided that 5 is the best option, because otherwise it goes past ***my*** Unbound setup. I hope Mozilla/Firefox will fix the two bugs linked above, so I don't have to choose between DNS under my control vs encrypted SNI.
*`network.trr.early-AAAA``true` to hopefully prefer IPv6
*`network.trr.uri` for the actual resolver address, e.g.
`https://dns.quad9.net/dns-query` or `https://149.112.112.112/dns-query` (removes the need for `network.trr.bootstrapAddress` and allows `network.trr.mode``3`?) or
* You can confirm TRR working by visiting `about:networking#dns` where
you should be seeing DNS cache of Firefox and a lot of `TRR: true`.
* Quad9 became my preferred resolver through anxiety about other options
being small (and possibly more likely to go down) or commercial while
Quad9 is non-profit organization and 2019-03-20 apparently the default
fallback resolver of dnscrypt-proxy (at least in Debian).
* Quad9 while having filtering of malicious domains should be easy to figure
out as the problem if something doesn't work on my computers as due to the
previously mentioned bug I am mainly using it on Firefox.
* [While investingating how Android 9 Private DNS works, I also wrote a DNS provider comparsion here]({% post_url blog/2019-07-11-android-private-dns-in-practice %})
#### SSDs
This information is from [Arch Wiki on Firefox tweaks](https://wiki.archlinux.org/index.php/Firefox/Tweaks)
*`browser.cache.disk.enable` to `false` to only cache to RAM.
* (`browser.cache.memory.enable` to `true` which should be default)
*`browser.sessionstore.interval` to `600000` in order to only store open session every ten minutes (instead of 15 seconds) in case of crashes.
* alternatively `browser.sessionstore.resume_from_crash` to `false` to not store the session data for crash recovery at all. I think this may be the more healthy option with all the information flood and dozens of tabs.
Why?
> Every object loaded (html page, jpeg image, css stylesheet, gif banner) is saved in the Firefox cache for future use without the need to download it again. It is estimated that only a fraction of these objects will be reused, usually about 30%. This because of very short object expiration time, updates or simply user behavior (loading new pages instead of returning to the ones already visited). The Firefox cache is divided into memory and disk cache and the latter results in frequent disk writes: newly loaded objects are written to memory and older objects are removed.
> Firefox stores the current session status (opened urls, cookies, history and form data) to the disk on a regular basis. It is used to recover a previous session in case of crash. The default setting is to save the session every 15 seconds, resulting in frequent disk access.
and this is the reason why Firefox is at times accused of killing SSDs.