One part of the filter is attached to our default DNS server, and it looks up novel domains to see if they ought to be blocked. This lookup is slow.
Launching Chrome first thing in the morning takes quite a while, because our DNS server/filter blocks until those 3 random domains fail to resolve.
I really wish that I could turn it off, although admittedly our "network" is barely compliant with protocols in the name of filtering.
--dns-server=8.8.8.8
But it also looks like it doesn't always work (never worked, or is a special development mode only feature). Here is the bug report on it:(thanks for pointing this out; it wasn't what i was suggesting - i had assume the laptop was the op's own).
Wondering why Chrome is not simply using the Google resolver on 8.8.8.8. Of course this would yet be another band aid that breaks the internet even further, when applications start implementing their own network stack.
Not a problem for me, but I'm sure a lot of people would complain.
This just seems like adding duck tape to a self-inflicted problem.
(Not that having chrome use 8.8.8.8 is a good idea either. As is often the case with hacks, either on its own works fine but the two interact poorly.)
That being said, it very well maybe time to find a better alternative :).
As for DNS, it has a very elegant hierarchical delegation system, such that organizations can control their own while also ensuring that all the world's DNS servers give consistent answers (and if you need short names for your internal systems on your own network, that's what search paths are for).
Roaming devices might make it a bit more tricky, but some integration with OS connectivity-detection and location-awareness might help there, as well as an improved UI for actually setting it.
A really short-term bandage would be to use some less identifiable DGA, but that opens the possibility of future malware using the Chrome DGA to pretend to be legit.
Is there a technical reason why 'foo.invalid' and 'bar.invalid' (and any of the other restricted dns names) couldn't be used instead?
I suppose that the ISPs could serve those correctly to make the tests pass, but what would they gain from it except even more user irritation?
Anything I'm missing?
A flag wouldn't be enough, because Chrome also learns what records it returns when it lies, so it can recognize other lies.
These "random" requests look almost exactly the DGA for Murofet- a Zeus variant. This has caused some problems for us (and other vendors, I would assume) in the form of massive numbers of false positives. In short, it's been kind of a PITA.
I wish they wouldn't do this, but it is definitely a tough problem to solve and I can't think of a better approach off of the top of my head. The ultimate culprit is the ISPs that return an A record for a DNS request that really should return NXDOMAIN. These ISPs are essentially breaking the Internet, and we're all just scrambling to put band-aids in place to get it to work again.