Brad Fitzpatrick

"The dot notation is a placeholder. Please ignore syntax for now." - Brad Fitzpatrick

"The dot notation is a placeholder. Please ignore syntax for now." - Brad Fitzpatrick

"Hey, I recognize that picture! I made that in 2004... http://danga.com/words..." - Brad Fitzpatrick

"Here's a rule of thumb: forget for a second that it's a package main, and ask yourself what you would do if it were a library. Stupid little detail? Lowercase. Stable API? Uppercase. Bonus: easy to migrate to be a library later. But even if not, the case is a good signal even within package main about which functions are intended for use and which exist just to help implement the uppercase ones." - Brad Fitzpatrick

"I find it's more common for people to hit this with maps, like: http://play.golang.org/p..." - Brad Fitzpatrick

"I didn't dispute that. But I think it's possible to write correct code, too." - Brad Fitzpatrick

"Can any of these people making Bitcoin sites write code? Shit. I should make a Bitcoin site." - Brad Fitzpatrick

"Riding the bus for 2 hours a day, frustrated at how terrible the Android build system was." - Brad Fitzpatrick

"We did not. It's not our job to make search work well for Go. It's search's job to work well for anything." - Brad Fitzpatrick

"There are many important production Go servers that would disagree with that statement." - Brad Fitzpatrick

"Or use Go and goroutines instead: all the joys of readable, top-down code, with the memory efficiency of events, and actually able to saturate all your CPUs (unlike, say, node or other single-threaded event systems)" - Brad Fitzpatrick

"Previously the poll server stuff (epoll or whatever) was implemented entirely in the net package, in pure Go. When they were blocked on a Read or Write, they'd detect that, register with the pollserver (pure Go), and wait on a channel for when they should try again. So then a goroutine would wake up when epoll_wait returned, process the events, tell all the registered waiters (sending on channels), and then they'd be runnable (since they received on their channel) and try again. Later that was extended to good effect to have N poll servers (one per CPU, up to some limit). It all worked very well, but the goroutine scheduling still had one global lock and didn't have any CPU affinity. The scheduler is now much smarter and has fewer locks, and the epoll_wait stuff is now integrated right into the scheduler, without ping-ponging around different goroutines like before." - Brad Fitzpatrick

"It's not fully precise yet. The stack precision isn't done, and some types of objects on the heap can still confuse it and be treated conservatively. But it's more precise." - Brad Fitzpatrick

"Yes, unless there's some major problem and it's reverted, but that's unlikely. The problem would probably just be fixed instead." - Brad Fitzpatrick

"Go has always used epoll for goroutines blocked on the network. But now the goroutine scheduler itself is aware of epoll (or kqueue, etc), so goroutines are run quicker and scheduled onto a more ideal processor than before, and with fewer locks involved." - Brad Fitzpatrick

«For example, “Hi, I’m Mxyztplk” is a grammatically valid sentence if your name happens to be Mxyztplk. (Come to think of it, it’s just as grammatically valid if you’re lying.) Clearly, it doesn’t make sense to count every string that starts with “Hi, I’m ...” as a separate sentence. To a normal English speaker, “Hi, I’m Mxyztplk” is basically indistinguishable from “Hi, I’m Mxzkqklt”, and shouldn't both count.» - 9000

It's probably still lower than that though, because just random words would not form a reasonable sentence. Most of these sentences would be practically as useful as "Hi, I'm Mxzkqklt", so it's important to figure out what you're really counting. - Ещё быстрей Гонсалес

The bikes taking up parking slots? Oh yes. - 9000

"I think that only happens on ARM. From the mailing list: Thanks to andrey mirtchovski here is the reformatted Raspberry Pi data, go version devel +ddb9e6365e57 Wed Feb 20 15:35:27 2013 -0800 linux/arm vs. go version devel +e00da5e201e5 Fri Feb 22 17:16:31 2013 -0800 linux/arm benchmark old ns/op new ns/op delta BenchmarkBinaryTree17 131488202467 112637283111 -14.34% BenchmarkFannkuch11 61976254131 61972329989 -0.01% BenchmarkGobDecode 424145307 383073401 -9.68% BenchmarkGobEncode 115032849 120332484 +4.61% BenchmarkGzip 13868472766 13493855517 -2.70% BenchmarkGunzip 1728657231 1680585482 -2.78% BenchmarkJSONEncode 2374700526 2377737536 +0.13% BenchmarkJSONDecode 5110739064 4877799205 -4.56% BenchmarkMandelbrot200 96515515 96478503 -0.04% BenchmarkParse 216403777 214091185 -1.07% BenchmarkRevcomp 192566195 196115543 +1.84% BenchmarkTemplate 6090635318 5572935111 -8.50% benchmark old MB/s new MB/s speedup BenchmarkGobDecode 1.81 2.00 1.10x BenchmarkGobEncode 6.67 6.38 0.96x..." - Brad Fitzpatrick

"No. That's probably about how much they currently cost to manufacture without setting up whatever assembly lines and tooling and other economies of scale to bring the price down to a level which market research says people will actually like / buy at. Also, you start working on hardware when it's new and expensive. It gets cheaper over time. By the time they care about selling them, the hardware will be cheaper and the economies of scale will be in place. (Disclaimer: I work at Google, but have nothing to do with Glass... this is all guesses, based on other hardware I've seen be made...)" - Brad Fitzpatrick

"Also: https://code.google.com/p..." - Brad Fitzpatrick

Wow, they internally introduced bound methods. I hope currying comes next :) #golang - 9000