Disable gaia in brave-core

[jumde]

On startup, Brave gets a list of the google accounts from gaia, which connects to accounts.google.com. Disable gaia to avoid these connections.

[diracdeltas]

could you list which domains are from gaia? i'm seeing some which might not be

[jumde]

On v 0.53.0. Here is the list of domains that I see on connect:

- accounts.google.com
- clients2.google.com
- ssl.gstatic.com
- update.googleapis.com
- googleapis.com
- clients2.googleusercontent.com

Here is a set of domains from gaia: https://cs.chromium.org/chromium/src/google_apis/gaia/gaia_urls.cc?q=gaia&dr=CSs

Issue to prevent connection to other google domains is tracked here - #663

[diracdeltas]

can this issue be fixed by setting the gaia URL constants to about:blank?

[jumde]

That might be a good solution, there may be side effects though. I can take a shot at this.

[jumde]

No connections are made to the gaia domains on browser-laptop on startup

[notDavid]

fyi, on 0.55.2 i see (after installing, on first startup, without any browsing):

www.gstatic.com
ssl.gstatic.com
www.googleapis.com
updates.googleapis.com
clients2.google.com
accounts.google.com
clients2.googleusercontent.com

(See screenshot here)

[bbondy]

New change landing in
brave/brave-core#512

Gaia URL will effectively always go to no-thanks.invalid which is a non resolve-able invalid host.
I'll keep this issue open for now though so we can work on seeing if we can not even compile the gaia code.

[LaurenWags]

Using

Brave	0.55.14 Chromium: 70.0.3538.54 (Official Build) beta(64-bit)
Revision	4f8e578b6680574714e9ed3bb9f02922b4dde40d-refs/branch-heads/3538@{#937}
OS	Mac OS X

This is what I see with LittleSnitch (first run, no browsing, using steps found here: #514 (comment))

[diracdeltas]

@LaurenWags i think those are all for extension updates (PDFJS), not Gaia related.

cc @tomlowenthal

[tomlowenthal]

Per discussion in slack, we're planning to serve PDFJS from the go-updater server rather than letting the browser connect to Google.

[tomlowenthal]

Moved this PDFJS sitch to its own issue: #1669 (nice).