Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
xtacy
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
31 ms
·
271.
▲
by
xtacy
14y ago
Earlier discussion on the same topic: http://news.ycombinator.com/item?id=4433142
272.
▲
by
xtacy
14y ago
That's very useful. I wish shells like bash/zsh had this instant auto-completion!
273.
▲
by
xtacy
14y ago
I really like the "instant" auto-completion. Can someone throw light on how it was done?
274.
▲
by
xtacy
14y ago
Hearing some fan noise, I immediately checked htop to see if any process is hogging the CPU. Nothing looked alarming, but turning off GDrive lowered CPU usage. I am not sure what the root cause is.
275.
▲
by
xtacy
14y ago
I doubt it was done on purpose. I think it's what minifying does to your CSS files.
276.
▲
by
xtacy
14y ago
I use the following changes to the style sheet for some more colour contrast: /* Increase contrast on some arrows */ .T-I-ax7.T-I-JE .T-I-J3 { opacity: 0.1 !important; } /* Make new chat highlight red instead of b
277.
▲
How SDN/OpenFlow can solve consumers' broadband woes
(gigaom.com)
2 points
by
xtacy
14y ago
|
0 comments
278.
▲
by
xtacy
14y ago
Open Networking Summit ( http://opennetsummit.org/ ) is a three day conference just for SDN/OpenFlow. If you're interested in seeing what companies are doing with OpenFlow, you should check it out. Incidentally, this year's summit starts t
279.
▲
Stanford study to try cold cash and social games to relieve rush hour traffic
(news.stanford.edu)
1 points
by
xtacy
15y ago
|
0 comments
280.
▲
by
xtacy
15y ago
Interesting visualization, but it may be a tiny bit misleading. My first thought was that the larger the horizontal bar, the more time spent in that function, but that's not the case. Rather, it captures the number of times the function
281.
▲
"230%" electrically-ultra-efficient LED from MIT
(wired.co.uk)
16 points
by
xtacy
15y ago
|
5 comments
282.
▲
StarCluster: Simplifying cluster management on EC2
(web.mit.edu)
3 points
by
xtacy
15y ago
|
0 comments
283.
▲
Safe Haskell: Including untrusted code safely to your trusted codebase
(scs.stanford.edu)
3 points
by
xtacy
15y ago
|
0 comments
284.
▲
Widescope: Leverage the Internet to illuminate and resolve the budget deadlocks
(widescope.stanford.edu)
1 points
by
xtacy
15y ago
|
0 comments
285.
▲
GDB tracepoint in a live running kernel
(code.google.com)
2 points
by
xtacy
15y ago
|
0 comments
286.
▲
by
xtacy
15y ago
Not many graduate students have really substantial industry experience when they start. Internships help a lot, provided the goal is to learn (and not be pressurised to publish papers even there. :)) It was probably unfortunate that you d
287.
▲
The TTY demystified
(linusakesson.net)
5 points
by
xtacy
15y ago
|
0 comments
288.
▲
Google seems to be broken on Firefox Nightly 13.0a1 (18 Feb)
(i.imgur.com)
1 points
by
xtacy
15y ago
|
0 comments
289.
▲
by
xtacy
15y ago
Another pedantic remark is that people use $1K instead of $1M to mean $1000. I think it's fine to use whatever it takes to get the message across.
290.
▲
Google wants to buy your browsing habits
(google.com)
8 points
by
xtacy
15y ago
|
0 comments
291.
▲
by
xtacy
15y ago
Nice collection. The weird concatenation, I suspect, is a part of the grammar. It's frequently used in C code as well, when the format string is too long to fit in one line. printf("%d ... blah blah blah" "%f ...", 1, 1.2
292.
▲
Nicira: A New Net
(nicira.com)
1 points
by
xtacy
15y ago
|
0 comments
293.
▲
by
xtacy
15y ago
Does it have integration with Dropbox or some other file sharing service? That, IMHO, would be a killer feature!
294.
▲
by
xtacy
15y ago
Good list of useful configuration options. I'd also like to add the "Compression yes" option that you can add on a per-host basis, and this could save some bytes sent over the wire. To see how much it saved, invoke with verbose ("ssh -v")
295.
▲
by
xtacy
15y ago
http://kernelnewbies.org is an excellent resource to get you started. They recommend starting here: http://kernelnewbies.org/KernelJanitors . I would also recommend subscribing to the KernelNewbies mailing list, to participate and as
296.
▲
by
xtacy
15y ago
Interestingly, this allows Google to measure the performance of a website from users' perspective by simply measuring the time it takes for a redirect. That information can be very helpful to see how the web works for everyone.
297.
▲
by
xtacy
15y ago
This is what AQM schemes like RED/ECN do!
298.
▲
by
xtacy
15y ago
Interesting. I've been looking into the Stephen Savage paper faking-acks part as well for a while now, especially in the context of hosted services in a shared infrastructure (cloud!). My hunch is that it can cause a lot of problems, bu
299.
▲
by
xtacy
15y ago
Shaping the link at the source works as long as you're the only computer along the bottleneck link. Once there are varying number of other sources competing for the bottleneck, determining the "rate" is hard (which is what congestion cont
300.
▲
by
xtacy
15y ago
On a related node, this page also has an interesting visualisation of the effects of cache misses on program performance: http://igoro.com/archive/gallery-of-processor-cache-effects
More ›