Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
_flux
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
87 ms
·
871.
▲
by
_flux
6y ago
Maybe 3 fingers more often than 2.
872.
▲
by
_flux
6y ago
I like Elecom HUGE and I've bound the rightmost button (so the fifth) so that when holding it the trackball works as vertical and horizontal scroll wheel. Works great for scrolling large areas, though there's still the regular scr
873.
▲
by
_flux
6y ago
But then at some point a breaking change in the backing service could be introduced, thus breaking pip2. Distros providing pip2 as part of their LTS will probably backport such changes.
874.
▲
by
_flux
6y ago
That may be the case now, but Facebook did start requiring Facebook accounts for their Oculus VR headsets. Much to the dismay of many. Maybe it's the test case FB is monitoring to look how to move ahead with their other acquisitions. L
875.
▲
by
_flux
6y ago
Surely you could put all that software on the expensive PC?
876.
▲
by
_flux
6y ago
They are (were?) hosted partially on Amazon, which should be able to scale quite nicely. If the system is built for it, that is. Though on 2018 there was some noise from Amazon about stopping their hosting due to practice called "Domai
877.
▲
by
_flux
6y ago
So I guess this calls for a led monitor project!
878.
▲
by
_flux
6y ago
So I picked the very first mems microphone datasheet my ddg gave me: https://www.st.com/resource/en/datasheet/mp34dt06j.pdf ..and it says 10 ms max turn on time, so 100 Hz. Of course, a typical or minimum tur
879.
▲
by
_flux
6y ago
https://github.com/dlundquist/sniproxy does not yet support ESNI. But might supporting ECH be completely infeasible? So for those unfamiliar, sniproxy is able to forward traffic to virtual hosts with their own encrypti
880.
▲
by
_flux
6y ago
This is what OCaml does: you can annotate function calls with [@tailcall] to result in compiler diagnostics if the call in fact isn't in tail call position.
881.
▲
by
_flux
6y ago
I guess it depends what you need and what is "secure and anonymous". Matrix probably leaks a ton of metadata. In practice it works by each device having their own encryption key and then those devices are bound together with a cro
882.
▲
by
_flux
6y ago
Well, matrix can to multi-device usage great and has E2E, so I don't see why that would be a hard requirement or even any kind of requirement.
883.
▲
by
_flux
6y ago
So regardless the amount of work it involves and how few people use them, e.g. 6-year old devices should always be supported? With PC the compatibility is quite a success story, but hardly the standard.
884.
▲
by
_flux
6y ago
Difficult to compare. I mean high CPU usage is good, it gets its job done faster, right?-) At least it is able to use multiple cores for it. Just started a backup and it seems the CPU usage stood still at at about 400%, varying from 90% to
885.
▲
by
_flux
6y ago
I think Kopia should deal fine with this, though I don't think I have many such large files that I modify. The reason why I think it's fine is that it doesn't really try to do "delta diffs" at all. Instead is uses a
886.
▲
by
_flux
6y ago
It would be nice to see kopia.io in the comparison. Kopia is a backup software based on the similar concepts as other modern apps (infinite increments, rolling-hash based deduplication, immutable blobs, encryption) but has some things going
887.
▲
by
_flux
6y ago
Unless daily work datasets are huge, you would be hard pressed to find a backup system that is unable to pull that off, though it's unclear from your description when you consider your backup to start. My multi-daily backups (always in
888.
▲
by
_flux
6y ago
I've used this extension for doing mass closes: https://addons.mozilla.org/en-US/firefox/addon/sorted-tabs/ As a bonus you can arrange URLs by name, url, or time next to each other, so it's usu
889.
▲
by
_flux
6y ago
Alt-F4 closes the entire browser, and indeed Ctrl-F4 closes the current tab. (Just tried it out, I've always used Ctrl-W for that.)
890.
▲
by
_flux
6y ago
In this case you could do it with GPT-x as well, if you have access to the learning material (so the OpenAI could do it when providing the model as a service): grep the learning data after-the-fact. No idea if it would be feasible to integr
891.
▲
by
_flux
6y ago
Amusing alternative: for model-as-a-service kind of things there could be blacklist or another neural network to handle the banned cases. "I'm afraid I cannot tell you that, Dave"
892.
▲
by
_flux
6y ago
Rather like the corpse of Pebble, after it killed itself by mismanagement.
893.
▲
by
_flux
6y ago
Email or the phone networks aren't global monopolies, are they? They even got text messages working globally—no idea about MMS though. So maybe there's a way to make a globally non-island IM that works in the Internet as well; or
894.
▲
by
_flux
6y ago
Consider adding https://github.com/kellyjonbrazil/jc .
895.
▲
by
_flux
6y ago
..not sure if a plug but this does exactly that: https://github.com/kellyjonbrazil/jc
896.
▲
by
_flux
6y ago
If it wasn't about money, surely they could have a second index for old code? And then asynchronously query both, filling the fast results in first. Or maybe have a keyword for searching old code, or search if current results seem bad,
897.
▲
by
_flux
6y ago
I think you missed that the speed limit would affect uploads after transferring 500G _per month_, not total? So in a year you would be able to store 6 terabytes. This would probably work out fine, because e.g. typical use of cloud storage i
898.
▲
by
_flux
6y ago
Speeding might get you through the lights before they turn red, though. A bad luck red sequence might altogether be avoided with some speeding. You might include "old green" in this as well, different regions have a different time
899.
▲
by
_flux
6y ago
Other suggest revenue source; but I think it's simply because it's easiest to enforce and least prone to personal judgement given it's based on strict numerical limits. Unsafe driving (braking distance, crossing lines, holdin
900.
▲
by
_flux
6y ago
Might be bigger than you think. Apparently people who use Realiser A16 have experienced the very same effect: once you fool the brains that speakers are stationary, you get a much more convincing effect about actually being in a studio. But
More ›