Now that's some wifi!

This is the new Network Box 2.0, my main project for the last 2 years or so.  How time flies!

(People are still trying to figure out why the upstream is weirdly slower than the downstream.  Latest theory I heard is it might be something about ring buffer replenishment.)

Ha.  Didn't see this last time around.

http://www.jwz.org/blog/2014/04/the-awful-thing-about-getting-it-right-the-first-time-is-that-nobody-realizes-how-hard-it-was/

I don't understand what's so hard about eventual consistency.  If any of your databases get out of sync, just wipe them and reinstall. Consistency. Done.

Neat.

https://www.opendns.com/about/innovations/dnscrypt/

Via an angry customer.  But hey, good information can come from all directions :)

I first knew I had a problem when I got an invitation to a presentation called "Quantized Congestion Notification" and got excited.

Co-worker: "That list of wifi projects your manager made is basically just propaganda about how much work you've been doing lately."

Me: "Um, well, actually that's a long list of projects that we haven't done yet."

Another co-worker: "It's getting to be a problem.  The list of things Avery hasn't done keeps getting longer and longer."

Yeah, pretty much.

I got a 'peer bonus' today.  I'm now suspicious that it was motivated by the same thing that motivated those people to give Obama a Nobel Peace Prize before he actually started work.

"Ha ha, suckers, I'm not falling for that one!" I'm sure he didn't say out loud.

From a presentation by Srikanth Sundaresan.  In short: latency matters, a lot.  Things stop getting better after 16 Mbps or so.

He claims that DNS prefetch (prepopulating a local DNS cache with popular names) can speed things up by about 10%.

I feel like someone had to lie or misinterpret or fudge something to come up with 59% of Americans having access to 100 Mbps internet.  But I don't know for sure.

http://arstechnica.com/business/2014/12/100mbps-internet-available-to-59-of-us-while-gigabit-still-at-just-3/

You can now link rust code to C code without a special runtime setup.

https://github.com/rust-lang/rust/commit/0efafac398ff7f28c5f0fe756c15b9008b3e0534

This is a very big attitude difference between Go (where mostly the approach is "rewrite the world in Go") and Rust (where it's becoming "we really need to interop or we'll fail).  Go has the advantage of having Google's muscle behind it, so "rewrite the world" will not automatically fail, but it's an interesting difference, for sure.

Asked interview candidate to design a radar system using software defined radio.  Candidate definitely did better than I did.  Unfortunately I still don't know how to design a radar system using software defined radio.

I think that says more about me than about the candidate.