5K was useful for pros who do 4K video editing, to have room for a UI around the video. But the next mastering size is 8K. I don't see what the point of 6K is except "it's one bigger".
The only bad thing about the (pre-Retina) Unibody era is those fucking Nvidia GPUs that were manufactured improperly and fall apart inside after a few years. Say what you will about modern Apple laptops and planned obsolescence, but Nvidia beat them to the punch.
We got ~68 years from 32 bits. Start with a 64-bit float, take 8 bits for the exponent, and you have 24 bits left for the fractional accuracy. 2^10 is about 1024, that's about a microsecond near the epoch, with 4 bits left over, so you get microsecond accuracy for another 64 years x 15, or 960 years. I'm going to guess that even millisecond precision is still sufficient as a general timestamp format. This is supposed to replace a timestamp of whole seconds, after all. The fractional part would be scaled from an internal clock in the computer, which might not have true millisecond precision anyhow. And there is no floating point math inaccuracy for whole seconds, because those are the integer portion of the float.
You would have to go a long time before those 24 bits get used up. Quick back-of-the-envelope is 68 years x 1024 x 1024 x 16, or just over a billion years, before the seconds start to skip. Not quite the heat death of the universe, but I think there are a few bigger problems before that happens. It's certainly a much better use of the extra bits than "duh just make it a long long" and waste 30 bits every time. All that would do is "merely" extend the problem out to ~300 billion years.
tl;dr: You haven't made a case for needing that much precision, over that time scale, in a timestamp.
NeXTstep (and by extension OS X) uses a double float seconds since 1970 in its NSDate class. This not only avoids the Y2038 (and Y10000) problem, but gives you microsecond accuracy for dates that are inside the current era. There is also no floating-point round-off error for any whole number of seconds because they are the base unit.
I'm in San Antonio, and hearing that they do not plan to abandon it (yet, at least) is good. Two-inch burial sounds really stupid, and Louisville is farther north, where they presumably get freezing weather much more than the handful of nights a year, with all-day freezing less than once a year on average, that happens in central Texas. I also live in a neighborhood with a back alley, so hopefully that will be cheaper to trench properly if they don't get to use the poles. But the ATT wires are underground along the alley too, so they'll have to pay some attention at least for the wires going across to individual houses. When I ran a BBS in the '90s and got a 3rd phone line, they trenched two 5-pair cables through my back yard between the junction boxes. I don't know how deep (I thought it was six inches), but there are places where it is already up to the ground.
When I lived in a neighborhood with no alleys in Austin many years ago, TWC re-cabled my block by digging a hole in every other backyard on one side, then used a hammer mole thingy to hammer a tunnel between them and then pushed conduit in behind it. They used the bottom half of an ordinary soda can that fit snugly over the end of the conduit to keep dirt out of the pipe, so it wasn't a tiny tunnel either. It was at least 12 inches down, maybe 24, so it should last forever.
Those are metal detector loops for vehicles stopped at an intersection, so they have to be that shallow. When road work is needed, they're the same people who put the loops in, so they can replace them at the same time. Nor are the wires carrying third-party data that is passing through, so if they fail, all that happens is people stopped at the intersection get inconvenienced.
Who are "they"? The real point of this controller is that it is a first-party controller with 1/8" jacks for everything. That separates the common part of a controller from the plethora of different controllers needed for a variety of disabilities. This means that the guts of the controller interface don't need to be duplicated in every specific unique custom controller, which are usually quite pricey because of low volume. $100 is cheap for something designed to have shit plugged into it. Also, if there is any kind of DRM in the controller (extra controllers are a lucrative market for console manufacturers, encouraging making them hard to replicate without a license), custom controllers become more difficult and expensive, often requiring a gutted retail controller be wedged in somewhere.
I discovered this recently because I started using FF only for playing Twitch video (to keep it out of my regular browser for various reasons). I found out that the display screen idle time-out will cause even the currently active tab, with FF as the front window, to not be considered active, and when the stream I was waiting for starts, it does not play. I can see how this is perfectly sensible as a default, but fortunately there is an override for those times when it is not desirable behavior, or when the detection has undesirable false positives.
I worked at Cisco back in the early 2Ks and I was surprised to discover that they had american.com (or maybe it was america.com), due to one of their many acquisitions. Of course the airline has its own 2-character domain, so they had no need for it.
I have a 3-character.com domain that I registered back in 2000. I used a random number function to generate each character, and the first one was unused. I tried a few more times and those were already taken. Back then in the first dot-com domain rush people were speculatively registering every 3-character.com domain out of the 26 x 37 x 37 (?) possible, and I just got lucky.
The Newton was crap until they got XScale CPUs. They started with 25MHz or so, and the XScale was like 250MHz. That was just when Newton started to get usable, so of course that was when they decided to toss it. Except for the Beeb and XScale days, until this decade, ARM had just never been a performance architecture, it was all about low power.
POWER wasn't inherently bad, the problem was that Motorola wanted to make low-power embedded chips (which ARM would later kill them in) hobbled with a slow front-side bus, and IBM wanted to make high-end workstation/mini-mainframe chips that needed liquid cooling (forget using them in a laptop).
And Intel had a lot of problems of their own, remember the Pentium 4? The Intel Mac came just as Core architecture started. Imagine the fun of having a Pentium 4 in a Mac.
A company called Levco made a "Monster Mac" board which could expand the 128K to 2 megabytes. A local guy was doing those upgrades, and I got one, probably in '86. It was pretty amazing having 2 megabytes of memory when 512K was the thing. I loved running that thing with a 1.5 megabyte ramdisk, and even wrote my own ramdisk program.
The upgrade basically entailed first upgrading the computer to 512K the hard way (clip the RAM chips, desolder their pins, toss chips in bin, add little board with extra mux chip), and certain chips were socketed (I think it was the 68000) to accomodate a daughter board full of sockets for 3x512K. They also made a SCSI interface which sandwiched under a ROM chip. And I had already gotten the upgraded Mac Plus ROMs, so I was ready with HFS while Hyperdrive users were seething trying to scramble to find some.
The upgrade also came with a little peizo fan that looked like a fat clothespin, with two vibrating plastic strips sticking up. And the guy also fixed my analog board when it inevitably failed. There was a non-polarized electrolytic rated at 12V or something, and when it failed, orange smoke came out of the computer. He replaced it with a big mylar thing with twice the voltage rating.
Adobe wanted way too much money for the license. It's one thing when you're making thousands of high-priced workstations, it's another when you're making millions of consumer computers. So Apple created Quartz, which was based on the PDF spec rather than Postscript. And it was also probably a good idea because my understanding is that DPS was too flexible (being a full programming language) and would let you do things that could hurt performance.
Note that "deodorant" and "antiperspirant" are two different things. Only the latter has aluminum in it. If you are not prone to excessive sweating, you should just use deodorant.
Slashdot Mirror
God wrote in Lisp code
I guess that means we know who writes in Perl.
5K was useful for pros who do 4K video editing, to have room for a UI around the video. But the next mastering size is 8K. I don't see what the point of 6K is except "it's one bigger".
The only bad thing about the (pre-Retina) Unibody era is those fucking Nvidia GPUs that were manufactured improperly and fall apart inside after a few years. Say what you will about modern Apple laptops and planned obsolescence, but Nvidia beat them to the punch.
Also glued-in keyboards that feel horrible and die at the end of the warranty period.
I got to see the second half of Rampage. It was surprisingly good for such a throwaway idea.
Don't worry, Ariane 6 will surely take over its throne.
They're not our crowd!
hell even the core elements of Taoism preach this.
I'm pretty sure that's Confucianism. People seem to get them confused as much as they do patents, copyrights, and trademarks.
precision varies with distance from the epoch
We got ~68 years from 32 bits. Start with a 64-bit float, take 8 bits for the exponent, and you have 24 bits left for the fractional accuracy. 2^10 is about 1024, that's about a microsecond near the epoch, with 4 bits left over, so you get microsecond accuracy for another 64 years x 15, or 960 years. I'm going to guess that even millisecond precision is still sufficient as a general timestamp format. This is supposed to replace a timestamp of whole seconds, after all. The fractional part would be scaled from an internal clock in the computer, which might not have true millisecond precision anyhow. And there is no floating point math inaccuracy for whole seconds, because those are the integer portion of the float.
You would have to go a long time before those 24 bits get used up. Quick back-of-the-envelope is 68 years x 1024 x 1024 x 16, or just over a billion years, before the seconds start to skip. Not quite the heat death of the universe, but I think there are a few bigger problems before that happens. It's certainly a much better use of the extra bits than "duh just make it a long long" and waste 30 bits every time. All that would do is "merely" extend the problem out to ~300 billion years.
tl;dr: You haven't made a case for needing that much precision, over that time scale, in a timestamp.
Some websites still don't seem to be designed with smartphones in mind.
#secondworldproblems
NeXTstep (and by extension OS X) uses a double float seconds since 1970 in its NSDate class. This not only avoids the Y2038 (and Y10000) problem, but gives you microsecond accuracy for dates that are inside the current era. There is also no floating-point round-off error for any whole number of seconds because they are the base unit.
If you think that's bad, you should read about all the dangers of Di-hydrogen Monoxide.
I'm in San Antonio, and hearing that they do not plan to abandon it (yet, at least) is good. Two-inch burial sounds really stupid, and Louisville is farther north, where they presumably get freezing weather much more than the handful of nights a year, with all-day freezing less than once a year on average, that happens in central Texas. I also live in a neighborhood with a back alley, so hopefully that will be cheaper to trench properly if they don't get to use the poles. But the ATT wires are underground along the alley too, so they'll have to pay some attention at least for the wires going across to individual houses. When I ran a BBS in the '90s and got a 3rd phone line, they trenched two 5-pair cables through my back yard between the junction boxes. I don't know how deep (I thought it was six inches), but there are places where it is already up to the ground.
When I lived in a neighborhood with no alleys in Austin many years ago, TWC re-cabled my block by digging a hole in every other backyard on one side, then used a hammer mole thingy to hammer a tunnel between them and then pushed conduit in behind it. They used the bottom half of an ordinary soda can that fit snugly over the end of the conduit to keep dirt out of the pipe, so it wasn't a tiny tunnel either. It was at least 12 inches down, maybe 24, so it should last forever.
Those are metal detector loops for vehicles stopped at an intersection, so they have to be that shallow. When road work is needed, they're the same people who put the loops in, so they can replace them at the same time. Nor are the wires carrying third-party data that is passing through, so if they fail, all that happens is people stopped at the intersection get inconvenienced.
Who are "they"? The real point of this controller is that it is a first-party controller with 1/8" jacks for everything. That separates the common part of a controller from the plethora of different controllers needed for a variety of disabilities. This means that the guts of the controller interface don't need to be duplicated in every specific unique custom controller, which are usually quite pricey because of low volume. $100 is cheap for something designed to have shit plugged into it. Also, if there is any kind of DRM in the controller (extra controllers are a lucrative market for console manufacturers, encouraging making them hard to replicate without a license), custom controllers become more difficult and expensive, often requiring a gutted retail controller be wedged in somewhere.
Concorde was one charter flight away from doing exactly that.
Somehow I don't think they were ISO-9000 certified.
No need to worry, he put the backup on a Batman flash drive!
I discovered this recently because I started using FF only for playing Twitch video (to keep it out of my regular browser for various reasons). I found out that the display screen idle time-out will cause even the currently active tab, with FF as the front window, to not be considered active, and when the stream I was waiting for starts, it does not play. I can see how this is perfectly sensible as a default, but fortunately there is an override for those times when it is not desirable behavior, or when the detection has undesirable false positives.
I worked at Cisco back in the early 2Ks and I was surprised to discover that they had american.com (or maybe it was america.com), due to one of their many acquisitions. Of course the airline has its own 2-character domain, so they had no need for it.
I have a 3-character .com domain that I registered back in 2000. I used a random number function to generate each character, and the first one was unused. I tried a few more times and those were already taken. Back then in the first dot-com domain rush people were speculatively registering every 3-character .com domain out of the 26 x 37 x 37 (?) possible, and I just got lucky.
The Newton was crap until they got XScale CPUs. They started with 25MHz or so, and the XScale was like 250MHz. That was just when Newton started to get usable, so of course that was when they decided to toss it. Except for the Beeb and XScale days, until this decade, ARM had just never been a performance architecture, it was all about low power.
POWER wasn't inherently bad, the problem was that Motorola wanted to make low-power embedded chips (which ARM would later kill them in) hobbled with a slow front-side bus, and IBM wanted to make high-end workstation/mini-mainframe chips that needed liquid cooling (forget using them in a laptop).
And Intel had a lot of problems of their own, remember the Pentium 4? The Intel Mac came just as Core architecture started. Imagine the fun of having a Pentium 4 in a Mac.
A company called Levco made a "Monster Mac" board which could expand the 128K to 2 megabytes. A local guy was doing those upgrades, and I got one, probably in '86. It was pretty amazing having 2 megabytes of memory when 512K was the thing. I loved running that thing with a 1.5 megabyte ramdisk, and even wrote my own ramdisk program.
The upgrade basically entailed first upgrading the computer to 512K the hard way (clip the RAM chips, desolder their pins, toss chips in bin, add little board with extra mux chip), and certain chips were socketed (I think it was the 68000) to accomodate a daughter board full of sockets for 3x512K. They also made a SCSI interface which sandwiched under a ROM chip. And I had already gotten the upgraded Mac Plus ROMs, so I was ready with HFS while Hyperdrive users were seething trying to scramble to find some.
The upgrade also came with a little peizo fan that looked like a fat clothespin, with two vibrating plastic strips sticking up. And the guy also fixed my analog board when it inevitably failed. There was a non-polarized electrolytic rated at 12V or something, and when it failed, orange smoke came out of the computer. He replaced it with a big mylar thing with twice the voltage rating.
Adobe wanted way too much money for the license. It's one thing when you're making thousands of high-priced workstations, it's another when you're making millions of consumer computers. So Apple created Quartz, which was based on the PDF spec rather than Postscript. And it was also probably a good idea because my understanding is that DPS was too flexible (being a full programming language) and would let you do things that could hurt performance.
Note that "deodorant" and "antiperspirant" are two different things. Only the latter has aluminum in it. If you are not prone to excessive sweating, you should just use deodorant.