Well, running the software on dedicated hardware (or on a full blown VM somewhere on the cloud - as several gaming solutions do) and streaming to your phone still beats everything in term of battery life vs performance.
(Also, you can turn the PC on and off remotely no need to have it run all the time. That's the whole point of Etherwake or newer technologies for lights-out management like IntelME, IPMI, etc. those even provide the VNC remote access. But saddly often also provide tons of exploitable bugs.).
Enables you to also leave the complex tax running on a power-grid machine and completely disconnect it from the phone (Zorpheus mentionned Photoshop. But such a setup would make even movie rendering possible). Enables to switch device (quickly do a few manipulation over the smartphone, then switch the remote controlling to the tablet when in a more comfortable train, then finally switch to physicall keyboard and screen once you're back home).
automatically switch between speakerphone mode (with half duplex, higher volume, and optional video) and non-speakerphone mode
The screen turning on is not speakerphone mode.
For the specific trick of automatically boosting the volume whenever the phone is away from the users' face, though it's not a default option on most android phones, I'm sure you'll find apps on the Play Store and a dozen of users' patchs on more hacker community oriented OSes (HP/Palm webOS, Jolla's Sailfish, etc.)
What ? it's a standard feature on the phone's cell voice calling on nearly any smartphone I've had. Put the smartphone against your ear: the screen shuts down, so your cheek won't accidentally click on stuff Put the smartphone down: the screen light ups showing you a numpad (e.g.: so you can click on a number pad for number-driven menus a.k.a. "Press 1 if you are calling regarding ###") and/or an optional toggle to increase volume.
This is just providing a web standard, so a a webapp like SkypeWeb can mimmick the same behaviour.
Voice calls are typically an exemple where the proximity sensors is used to shut down the screen and touch sensors. (e.g.: If the proximity sensors senses that the whole face is against the screen, because the user is holding the phone against the ear instead of using earphones / bluetooth, then this is necessary to avoid the users' cheek clicking on random stuff on the screen).
Skype has a fully functionnal web app (web.skype.com) and on Linux, the new "native" app is basically just the web app wrapped together with a chrome browser. Such an app would be a prime candidate for using such API.
Now a single web standard can be used whenever the user hold a smartphone to make a call, instead of each wrapped web app requiring special plugins calling into whatever that peculiar platform uses. (be it web.app in Firefox on your smartphone, web.app wrapped in a browser on your ubuntu touch phone, perhaps even web.app wrapped in an android app)
you've got the wrong scale regarding proximity. it's not about the number of meters from the device to the user. it's more "is there something pressed against the screen ?" (like the user's cheek and ear in case of a smartphone). That's the thing that turn off the screen and (even more important) the touch screen while you're talking into the phone (that is : with the whole smartphone's body against your head. Not using earphones / bluetooth).
Web apps are a thing, and in some case, such as Skype, the web app (http://web.skype.com in this case) *is* the app. The new "native" Linux client is basically the web app, but packaged together with chrome, thanks to Electron framework. (and initially, wrapped together with the necessary plug-in to enable voice-chat, back when it used microsoft-edge-only api instead of webrtc. things have moved since and call work in bare firefox without any plugin and only webrtc).
(I haven't bothered to check, but I strongly suspect that the android app also shares code heavily this web app)
By providing an API to light and proximity sensors, it gives the web app a (web-)standard way to be able to behave like a normal phone app during calls and have the screen shut down so your cheek won't accidentally click on stuff. Thus SkypeWeb opened in Firefox will act the same as the regular app, and all of them could potentially use the exact same API (instead of the Android App, iOS App, Windows APP all using their system's specific sensors API and Firefox not even being able to).
Same way of thinking could be applied to any other on-line voice calling platform's web app, or even "native" for the cases where "native" is just a wrapped up web app.
The problem, is that some rogue script on some mailicious website/ad could abuse it: e.g only do nefarious things such as mining cpu-cryptocoin only when the user isn't looking. So these API will go the same way as all other sensitive API such as Location/GPS API : diabled by default, let user only enabled them on the sites where it's genuinely needs (Location: e.g. on google maps).
Now, Intel also did never manage to come up with anything x86 that was suitable for a smartphone.
Worse. They did never manage to come up with anything specifically running the x86 instruction set that was suitable for a smartphone. They used to have a decent Intel-manufactured CPU running ARM instruction set, but somehow managed to abandon the market and sell it off, just at the time when ARM is getting even more relevant thanks to smartphones, routers and IoT.
Search for "Intel StrongArm" and "Intel XScale".
Note that, according to Wikipedia, Intel is still in possession of ARM license that they acquired when bought StrongArm. So even after selling XScale out to Marvell, they could still start a new line of ARM core *now*, after having come to realization that the Atom doesn't scale down as much as they would have liked (isn't that well suited for smartphones and routers) and its x86 compatibility makes absolutely not sense in those markets (Seriously, nobody is going to run legacy Windows code on a smartphone)
AMD did not even try, because they understand CPUs and knew this architecture is not suitable for that field. But they went one step farther: They have server processors that include ARM cores.
I'm still hoping that, next to the ARM light-weight servers that they are targeting, these ARM cores will eventually also evolve to some high range phablets and dev boards.
The problem is that only high-end devices are just getting H.265 support
(which is also due to the patent mess and thus nobody being in a hurry to jump into the bandwagon)
If the industry keeps coming out with new codecs at this rate,
If you look, "the industry" is more or less grouping around two entities : - the MPEG which still tend to design codecs the old way (file patents and monetize through licensing) - and the Alliance for Open Media, where basically any industry member that has anything to do with video in their business is represented (the whole chain from the camera to the mobile device receiving the stream seem to be represented) with a completely different approach to financing it (these are video companies. they earn money from video any way : be it selling hardware, services, etc. they don't need to sell AV1, they only need to get rid of the licensing madness of h265. Making an open/free codec makes entirely sense for them)
So we're not bound to see dozens of new codecs, we're basically only expecting 2 : - whatever MPEG comes up with after H265 - AV1 given the list of members behind AV1, it's bound to be supported by most software and hardware pretty fast.
we are going to have to adapt our chips to be more flexible, perhaps even programmable in the hardware decoder area
Congratulation, you've successfully described a compute shader. More seriously : GPU have flexible blocks - the compute shaders available to Vulkan and OpenCL. They might not be as efficient as a dedicated core, but they are deffinitely better than a naked CPU core. AV1 is on purpose designed in a way to be easily implemented on a GPU, and decision are taken in this favor.
H.264 and VP8 is easy to find these days,
Yes.
VP9 and H265 is slowly but surely coming,
h265 is *very slow* at coming, mostly due to the patent minefield making manufacturer less in a hurry to support it.
releasing a brand new codec today will take 5 years to get it in the majority of high-end chip fabs and another 5-or-so years to go mainstream with at least 15-20y more years of having to have both available.
Unless the chip manufacturer are part of the process. Which is the case (ARM, AMD, Nvidia, Broadcom, etc.) They are considering hardware implementation while the AV1 is being designed and contributing appropriate feed-back. - compute shader code will be available at codec release time (you'll be able to have decent performance on smartphone on day 1). - the manufacturer plan to have chips ready within one year. - means by 3 year (counting the current 2 year churn) there will be a lot of smartphone native-capable on the market.
You assume Apple and Google will allow software decoding of AV1, which is extremely bad for battery life.
Apple : No, they wont. They have high stake in H265 patents. (I am actually surprised that there are part of the alliance)
Google: Yes, they will. AV1 is also designed to be easy to implement in hardware and in GPGPU acceleration. Means that, there will be some implementations on whatever is closest to a OpenCL / Vulkan combo available on the hardware. So even before AMD, Nvidia, and the other hardware manufacturer of the alliance start shipping dedicated AV1 hardware on their GPU, the GPU will already be able to offload a significant bit.
Your battery won't suffer as bad as if everything was 100% decoded by the ARM CPU core.
For any given video, it will recommend a range of other similar videos which by definition must be a bit more radical or a bit less radical. If you keep clicking the more radical ones, of course you will slowly gravitate up the radical tree. How could it be otherwise?
It goes a tiny bit further :
- it's been already studied and proven (on oldschool TV) that more extreme content (specially more frightening) increases viewer engagement.
- and engaged viewers will bring more revenue by selling their eyeballs to advertisers.
- (this happens even more on private channels than on public TV).
- thus TV channels, specially news casts, tend to gravitate toward more
The AI neural net behind Youtube recommendation just simply "independently rediscover" what's been studied regarding old school TV. (while being probably even less aware of it : during A/B tests the algorithm only notice that video on list A tend to increase viewer retention compared to list B and thus maximize ads exposition and revenu stream. it just happens that the videos on this list A are the most extreme due to what we already know of human psychology and past TV studies. The algorithm will eventually automatically build a chain of recommendation of increasing extremeness, because that's what works better for the result it tries to maximise)
The sad thing is that this has been also proven to increase the feeling of insecure.
So, yes, initially the youtube algorithm will show up a variety of similarily themed video recommendations, some of which "must be a bit more radical or a bit less radical". But eventually some of these recommendation will prove more popular and youtube will learn to show them more. Due to how human psyche works, those more successful videos *will be* a little bit more frequently the more radical ones. And thus youtube learns to show the radical more a bit more often (without even having the notion of what "radical" is, only that they are successful). And again, sadly due to how human psyche works, it will have a negative impact on viewers.
it's been discussed for quite some time that TV channels, specially their News, tend to over-dramatize news and cast dramatic negative light on the world. Basically, when creating emotion in viewer, their attention increases. This works even more with negative emotions (fear, etc.). This attracts more eyeballs to your channels, giving you more opportunity to sell these attentive eyeballs to advertisers and thus increase your revenue stream. In Europe, this is more prominent on private channels (mostly paid by ads) than on public channels (partially paid by taxes).
The thing is that, in practice, it has been proven to have an actual effect : in Europe, watching TV and watching news even more so, has been linked to causing an increased feel of "insecurity, danger, etc." This is despite the situation in Europe being much better and safer than before. Criminality rate is globally decreasing, but TV reporting thereof is on the rise.
The neural-network AI used by youtube to process recommendation has simply rediscovered on its own the same results as what was already found on TV : increase the emotional response of viewer by showing more extreme videos, you attract their attention and thus can sell more ads.
The AI doesn't even really have an actual concept of "emotional response" and "increased attention". The AI only notices that after recommending some video, revenue increase. If video B is shown after video A : more retention, increased ads revenu. If then video C is shown after video B : again more. AI the remembers to use the chain A -> B -> C, because that's what increases the parameter it is optimising for : eyeball time sellable to advertisers. But because of what we know from what was studied on old school TV, that will eventually mean showing more and more extreme videos, because that's what has been proven to work on human brains. This AI has simply "independently rediscovered this fact".
----
For shit and giggles, let's gice a source that's on Youtube itself (and even parodies the usual style of conspiratorial videos). Sorry it's in French, but it has English CC.
Interestingly, The Concorde stopped flying in 2003, the very year Skype launched.
Funny, I simply used Skype as a metaphor. Didn't think about the timing.
But that had zip to do with businessmen/women's reason for not using it. Business travel has not decreased. Did you know that millennials are the most frequent business travelers?
I'm not saying that business travel is decreasing. I'm saying that there's a very specific subtype of business travel ("This is a very important meeting, you need to go represent the company, now !!!") that require urgently transfering a business person as fast as possible on the premise. That specific type of *urgent* business meeting has decreased, and thus a small barely significant use-case for concorde has disappear. I'm not saying that business would save the concorde. I'm simply say that's about the single (rare) situation case where one might legitimately have an use of the ultra-sonic speed of concorde (but not making economical sense).
For everything else, it was basically a "showing off" project.
"2. Nuclear fusion will come next. It is just an engineering problem."
Funny how in the previous half-century we managed to solve every other "engineering problem", we weren't able to solve that one.
"It's just an engineering problem" == the science around it is basically understood. We don't need to discover or develop new science (as opposed to true anti-grav hover boards). But it is a ginormous project that will take decades just to build.
To take a space race metaphor : Once Sputnik is in orbit, it doesn't take a long strecht of imagination to think that humanity will also manage to put people on the Moon or built a giant international space station. But from the Sputnik point in time to the modern day's ISS, it took a ginormous amount to resources, money, time and engineering.
The 15 years time frame touted by the project is probably over-optimistic. But hey, we live in the "startup / kickstarter / etc." era. Where over promizing is an absolutely necessary first step to attract the attention of money sources.
We don't even have the Concorde anymore, how come that solved problem isn't even part of the landscape any more?
Solved problem/working solution DOES NOT mean "makes sense economically". Concorde for most of its life was a vanity project that never manage to make sens economically and look a long-term sustainable project. And that's even before the modern ultra-connected world removed even more of the few corner cases where Concorde might have been useful. (Basically, Skype killed the need to have business man able to cross the world within hours - and these 3 business men where the only out of all passengers that actually had a practical use of Concord) (Though nowadays, if it was still around, Concorde might successfully be marketed as an "exclusive" experience to a few ultra-rich 1%er with way too much money in their pockets).
In other words, all the/. who currently heavily criticize the economical sustainability of Tesla/Uber/etc. would probably thing that those are still perfectly long-term stable plans compared to Concord.
Fusion, like space colonies, is a fantasy for weak-minded tech nerds looking for a religious experience to come from technology.
Space colonies are technically feasible and limited to engineering. But such a vast problem requiring so many resources compared to potential use now, that nobody is ever going to thow the necessary money at it. You would need dozens of government putting all their resources together for a century or two to even hope boot strapping it. (Or the US divert a few percent of your war budget:-P )
It's basically a project on the same scale as the colonization of the world by the big maritime empires, except back then the empire were having high hopes to gain riches, whereas there aren't much economic incentives to invest into space colonization now. Vague hopes of mining helium 3 on the moon do not sound a credible enough excuse to lock major part of the budget of dozens of government over the next couple of centuries.
And scientific discoveries aren't a good enough incentive. It's currently way cheaper to send probes and robot to explore instead of the costly and complex project of putting humans in person there.
Sorry, Bill. You won't see commercial fusion power, asteroid mining, or space colonies.
Well it's not clear when fusion will make sense economically - might take 40 years, might take a century - or might end up not making sense at all.
But the current big project like ITER, are currently done for the sake of science, as well as vanity project to show off engineering capabilites and try to learn useful stuff out of it.
We might manage to get something out of ITER - at least learn a lot - though it might take a lot more resource and time than initially planned.
TFA's startup ? well... I wouldn't hold my breath.
Anyone know how many apps out there actually use SGX?
On Linux (and other open source OS, i.e.: the only distributions where code for nearly all the software is available and can be recompiled with a retpoline-enabled compiler, such as the recent GCC 7.3.1, and thus the only environment where there's any hope for spectre counter measures to be actually deployed): probably close do zero, anyway.
Not really. Not if both end stations have electric cars for rent that only need a battery range of 40 to 50 miles.
Which is actually a real-world situation in switzerland, with the biggest car-sharing cooperative (Mobility) also having cars available at trains stations, including electric cars in bigger cities (Renault Zoe - currently equipped with the smaller 125km range battery, progressively getting upgraded to the bigger 250km).
This is currently successful commercially.
no need for fast charging infrastructure either. Rail plus electric cars is a beautiful combination
Actually, due to how train work, you happen to have a fast charging solution available almost for free at the train station.
Cygwin is the WINE-like translation layer. WSL is about the NT kernel able to speak Linux API directly. (NT kernel is weird in that it can support several different API - used to also directly run OS/2 software without translations).
Clearly I/O is the issue, since compiling is very I/O intensive. {...} you will notice that the WSL is almost 100x slower than the bare metal version of Linux when compiling.
Because it doesn't matter if your kernel is able to directly serve IO request natively instead of needing translation... When those IO request are served by the usually terribly sucky NTFS code.
The main IO bottle neck (Windows' own awefull performance on anything filesystem related) is still there, no matter how many (insignificant) abstraction layers you removed above.
The major performance issues that remain are with I/O.
Because even if there's no translation happenning, you're still bound to the sucky NT Kernel filesystem drivers.
Otherwise it's actually pretty good, in some cases equal to or even slightly better than bare-metal Ubuntu performance.
Keep in mind that this has mostly to do with the way multi-processing is handled : - Windows suck at multiple processes, because creating a new context ( fork() ) is a horribly inefficient process. (Whereas on Linux, it's almost a free action thanks to CoW facilities in the virtual-memory subsystem). - As such when running a unix software using a software translation layer (like Cygwin), multi processing will suck. - That's why multi-threading is popular in Windows world : there's no context separation, everything is done in the came context.
- The NT kernel introduced a new concept called pico-thread which are much more light-weight than regular Windows process to setup. These aren't available in Windows, but gives a way to the NT kernel to provide extremely light-weight multi-processing to Linux ELFs. - Multi-processing works decently well on WSL (unlike Windows native apps, or Unix apps via Cygwin).
- But if you read the technical blogs at microsoft, you'll release that the managed to achieve pico-threads by throwing away some of the context isolation of actual multiple process. (There's a reason while picothreads aren't available for production Windows software)
- So basically in purely multiprocessing/multithreading benchmarks (e.g.: thinks running on OpenMP) WSL can even slightly beat actual real linux, because Microsoft threw a lot of safety and security out of the window. (It's great for testing software, but do not ever contemplate using WSL in production. It's only to test software before deploying on real Linux). - When benchmarks are mostly CPU oriented (e.g.: most of the media compression tests) - most of the CPU cycles are spent running the instruction to process the data, and they are the same no matter what OS they run on (i.e.: a cygwin compiled software would run just as fast, provided it was compiled with a similar version of GCC the optimize code the same way).
- Whenever a benchmark hits any other part of the NT kernel (example: file IO) the performance just completely collapses. It doesn't matter that there's no translation going on and that the NT kernel is directly service IO request it self, when that IO code just plain sucks. A deep overhaul of the NTFS code would be the only hope for WSL to suck a tiny bit less.
(3D API in theory could be an area where performance degradation won't be as significant : some manufacturer like Intel produce better Windows drivers than Linux, and other like Nvidia re-use basically the same drivers. But OpenGL is among the long list of Linux APIs that WSL is not supporting in its (very limited) subset).
Wine is a translation layer, it's a piece of user land software that maps incoming Windows API calls from native.EXE to the closest approximation achievable under Unix/POSIX systems (quite a lot is achievable actually).
WSL is NT kernel having multiple personnalities. In addition to be able to directly serve Windows API to.EXE, the NT kernel itself also able to serve (an extremely small subset of) Linux API to ELFs, including stuff that doesn't exist in Windows API and could not be done by a userland translation layer such as Cygwin (Windows API sucks at multi-processing, but the NT Kernel has recently gained pico-threads which makes it possible to support decent multi-processing for Linux ELFs, despite such a thing not existing in the regular linux world). (On the other hand you're still limited to the NT kernel's horrible IO)
(A long time ago the NT kernel used to do the same for OS/2 applications - run them directly without any translation in the middle).
And then cue in companies complaining how they now need much more CPU in their servers, because instead of blindly serving files in the clear, they need to secure traffic with HTTPS and check authorization to send data. And complaining how it costs more bandwidth because every single piece of data must be sent again over internet and can't be cached due to access control. And how some people will be capturing and storing the streams locally any-way.
Here is where the war is currently. Add in phones and tablets and other walled off closed ecosystems and you can see this is a bad road where freedom doesn't exist.
If data is kept locked on the corporation's cloud, do you even need the phone and tablet ?
You can just keep the software as web app, and have the users keep paying a recurring fee if they want to have the privilege of keeping to work with the data they left locked. (see Software as a Service, "Microsoft Office Online", etc.)
You invite someone over for dinner and they steal your cutlery and napkins.... what do you do?
The weird thing with digital media and where this metaphor breaks down, is that even if that someone stole your silver cutlery and silk napkins, next time you open the drawers of your kitchen, they'll be magically full again, and you can still invite someone else on your table with proper silverware.
The closest thing would be inviting someone for a private view of a master's painting in your living room. But the guy take a polaroid, so he can look at it whenever he wants (shitty quality of the analog compared to what digital media replication allows aside).
The main problem is that we still apply an absolutely out of date method to pay artists for creations - making the duplication the paying step, even if nowadays it's actually the simplest and cheapest step of the process.
The industry solution is to try to invent some magical snake oil that suddenly makes Polaroid not work in the presence of work of art.. Which is cryptographically impossible to achieve when the usual "Alice" and "Eve" are actually the same person.
The actually potentially successful solution would be trying to find better ways to pay artists to keep creating, while not taxing the simplest part. Patronage and crowdfunding seem to be potential parts of the solution.
Will probably use the "LE" (Low Energy) Bluetooth protocol (it's also under the Bluetooth umbrella name, but has nothing to do with classical Bluetooth, it'ss more some kind of ZigBee-like low power protocol that got phagocyted by Bluetooth).
Which is completely stupid in this context(*), but helps Apple add a bullet point on the marketing speech ("Low Power !") and helps them use a protocol (LEA) which isn't widely used by anyone else.
---
(*): Audio is continuous streaming which is not something LE excels at. LE is good for quick short burst message that only happen every now and then (like a health device giving a status : think heart beat monitors).
We already had this discussion back when Apple was spinning some marketing about having made LEA available to auditory aid equipment and thus being better than Android for people with hearing deffects.
But 30-40 years ago, the DCF77 standard started to become very popular.
Instead of having clock that still need to be set after a power outage or needs to be set after each daylight saving time change, you have a clock than can auto-adjust it-self hourly over the radio from an atomic clock in Germany, that still works across most Europe (i.e. within the long range radio's reception), is only a few seconds off (the ping time of the radio transmission and electronics processing), and has a notion of DST (so no need to manually move 1 hour forward or backward), and can even work on battery without anything plugged in (it's radio).
And it's German, so nobody will complain that it's tremendously over-engineered as a solution.
At some point in time in the past, TV stations radioemitter could use DCF77 as a source for the clock they broadcast over Teletext/Videotext pages.
But they fucked up because the "free shareware episode" here already contains the data for the extra "commercial episodes" and only a small switch gets flipped to enable continuing.
(Whereas, you got extra floppies with Commander Keen episode 2 and 3, or Wolfenstein Episodes 2 to 6, once you paid after testing the free shareware episode 1).
(Though I've read somewhere that Id *DID* make the same blunder with the shareware CD of Quake1. There was only a simple registration procedure that was supposed to unlock the full commercial game with extra episodes out of the shareware CD, the same way here the full Yakuza6 can apparently be unlocked out of free 36GB demo).
Slashdot Mirror
Well, running the software on dedicated hardware (or on a full blown VM somewhere on the cloud - as several gaming solutions do) and streaming to your phone still beats everything in term of battery life vs performance.
(Also, you can turn the PC on and off remotely no need to have it run all the time. That's the whole point of Etherwake or newer technologies for lights-out management like IntelME, IPMI, etc. those even provide the VNC remote access. But saddly often also provide tons of exploitable bugs.).
Enables you to also leave the complex tax running on a power-grid machine and completely disconnect it from the phone (Zorpheus mentionned Photoshop. But such a setup would make even movie rendering possible).
Enables to switch device (quickly do a few manipulation over the smartphone, then switch the remote controlling to the tablet when in a more comfortable train, then finally switch to physicall keyboard and screen once you're back home).
Read what I replied to more closely:
automatically switch between speakerphone mode (with half duplex, higher volume, and optional video) and non-speakerphone mode
The screen turning on is not speakerphone mode.
For the specific trick of automatically boosting the volume whenever the phone is away from the users' face, though it's not a default option on most android phones, I'm sure you'll find apps on the Play Store and a dozen of users' patchs on more hacker community oriented OSes (HP/Palm webOS, Jolla's Sailfish, etc.)
The phone's own dialer doesn't even do that.
What ?
it's a standard feature on the phone's cell voice calling on nearly any smartphone I've had.
Put the smartphone against your ear: the screen shuts down, so your cheek won't accidentally click on stuff
Put the smartphone down: the screen light ups showing you a numpad (e.g.: so you can click on a number pad for number-driven menus a.k.a. "Press 1 if you are calling regarding ###") and/or an optional toggle to increase volume.
This is just providing a web standard, so a a webapp like SkypeWeb can mimmick the same behaviour.
Voice calls are typically an exemple where the proximity sensors is used to shut down the screen and touch sensors.
(e.g.: If the proximity sensors senses that the whole face is against the screen, because the user is holding the phone against the ear instead of using earphones / bluetooth, then this is necessary to avoid the users' cheek clicking on random stuff on the screen).
Skype has a fully functionnal web app (web.skype.com) and on Linux, the new "native" app is basically just the web app wrapped together with a chrome browser.
Such an app would be a prime candidate for using such API.
Now a single web standard can be used whenever the user hold a smartphone to make a call, instead of each wrapped web app requiring special plugins calling into whatever that peculiar platform uses. (be it web.app in Firefox on your smartphone, web.app wrapped in a browser on your ubuntu touch phone, perhaps even web.app wrapped in an android app)
you've got the wrong scale regarding proximity. it's not about the number of meters from the device to the user.
it's more "is there something pressed against the screen ?" (like the user's cheek and ear in case of a smartphone).
That's the thing that turn off the screen and (even more important) the touch screen while you're talking into the phone (that is : with the whole smartphone's body against your head. Not using earphones / bluetooth).
Web apps are a thing, and in some case, such as Skype, the web app (http://web.skype.com in this case) *is* the app. The new "native" Linux client is basically the web app, but packaged together with chrome, thanks to Electron framework. (and initially, wrapped together with the necessary plug-in to enable voice-chat, back when it used microsoft-edge-only api instead of webrtc. things have moved since and call work in bare firefox without any plugin and only webrtc).
(I haven't bothered to check, but I strongly suspect that the android app also shares code heavily this web app)
By providing an API to light and proximity sensors, it gives the web app a (web-)standard way to be able to behave like a normal phone app during calls and have the screen shut down so your cheek won't accidentally click on stuff.
Thus SkypeWeb opened in Firefox will act the same as the regular app, and all of them could potentially use the exact same API (instead of the Android App, iOS App, Windows APP all using their system's specific sensors API and Firefox not even being able to).
Same way of thinking could be applied to any other on-line voice calling platform's web app, or even "native" for the cases where "native" is just a wrapped up web app.
The problem, is that some rogue script on some mailicious website/ad could abuse it: e.g only do nefarious things such as mining cpu-cryptocoin only when the user isn't looking.
So these API will go the same way as all other sensitive API such as Location/GPS API : diabled by default, let user only enabled them on the sites where it's genuinely needs (Location: e.g. on google maps).
Now, Intel also did never manage to come up with anything x86 that was suitable for a smartphone.
Worse.
They did never manage to come up with anything specifically running the x86 instruction set that was suitable for a smartphone.
They used to have a decent Intel-manufactured CPU running ARM instruction set, but somehow managed to abandon the market and sell it off, just at the time when ARM is getting even more relevant thanks to smartphones, routers and IoT.
Search for "Intel StrongArm" and "Intel XScale".
Note that, according to Wikipedia, Intel is still in possession of ARM license that they acquired when bought StrongArm.
So even after selling XScale out to Marvell, they could still start a new line of ARM core *now*, after having come to realization that the Atom doesn't scale down as much as they would have liked (isn't that well suited for smartphones and routers) and its x86 compatibility makes absolutely not sense in those markets (Seriously, nobody is going to run legacy Windows code on a smartphone)
AMD did not even try, because they understand CPUs and knew this architecture is not suitable for that field. But they went one step farther: They have server processors that include ARM cores.
I'm still hoping that, next to the ARM light-weight servers that they are targeting, these ARM cores will eventually also evolve to some high range phablets and dev boards.
or Spotify should buy SpotLister.
(See the vertical integration in radio markets nowadays).
The problem is that only high-end devices are just getting H.265 support
(which is also due to the patent mess and thus nobody being in a hurry to jump into the bandwagon)
If the industry keeps coming out with new codecs at this rate,
If you look, "the industry" is more or less grouping around two entities :
- the MPEG which still tend to design codecs the old way (file patents and monetize through licensing)
- and the Alliance for Open Media, where basically any industry member that has anything to do with video in their business is represented (the whole chain from the camera to the mobile device receiving the stream seem to be represented) with a completely different approach to financing it (these are video companies. they earn money from video any way : be it selling hardware, services, etc. they don't need to sell AV1, they only need to get rid of the licensing madness of h265. Making an open/free codec makes entirely sense for them)
So we're not bound to see dozens of new codecs, we're basically only expecting 2 :
- whatever MPEG comes up with after H265
- AV1
given the list of members behind AV1, it's bound to be supported by most software and hardware pretty fast.
we are going to have to adapt our chips to be more flexible, perhaps even programmable in the hardware decoder area
Congratulation, you've successfully described a compute shader.
More seriously : GPU have flexible blocks - the compute shaders available to Vulkan and OpenCL. They might not be as efficient as a dedicated core, but they are deffinitely better than a naked CPU core. AV1 is on purpose designed in a way to be easily implemented on a GPU, and decision are taken in this favor.
H.264 and VP8 is easy to find these days,
Yes.
VP9 and H265 is slowly but surely coming,
h265 is *very slow* at coming, mostly due to the patent minefield making manufacturer less in a hurry to support it.
releasing a brand new codec today will take 5 years to get it in the majority of high-end chip fabs and another 5-or-so years to go mainstream with at least 15-20y more years of having to have both available.
Unless the chip manufacturer are part of the process. Which is the case (ARM, AMD, Nvidia, Broadcom, etc.)
They are considering hardware implementation while the AV1 is being designed and contributing appropriate feed-back.
- compute shader code will be available at codec release time (you'll be able to have decent performance on smartphone on day 1).
- the manufacturer plan to have chips ready within one year.
- means by 3 year (counting the current 2 year churn) there will be a lot of smartphone native-capable on the market.
You assume Apple and Google will allow software decoding of AV1, which is extremely bad for battery life.
Apple :
No, they wont. They have high stake in H265 patents.
(I am actually surprised that there are part of the alliance)
Google:
Yes, they will. AV1 is also designed to be easy to implement in hardware and in GPGPU acceleration.
Means that, there will be some implementations on whatever is closest to a OpenCL / Vulkan combo available on the hardware.
So even before AMD, Nvidia, and the other hardware manufacturer of the alliance start shipping dedicated AV1 hardware on their GPU, the GPU will already be able to offload a significant bit.
Your battery won't suffer as bad as if everything was 100% decoded by the ARM CPU core.
For any given video, it will recommend a range of other similar videos which by definition must be a bit more radical or a bit less radical. If you keep clicking the more radical ones, of course you will slowly gravitate up the radical tree. How could it be otherwise?
It goes a tiny bit further :
- it's been already studied and proven (on oldschool TV) that more extreme content (specially more frightening) increases viewer engagement.
- and engaged viewers will bring more revenue by selling their eyeballs to advertisers.
- (this happens even more on private channels than on public TV).
- thus TV channels, specially news casts, tend to gravitate toward more
The AI neural net behind Youtube recommendation just simply "independently rediscover" what's been studied regarding old school TV.
(while being probably even less aware of it : during A/B tests the algorithm only notice that video on list A tend to increase viewer retention compared to list B and thus maximize ads exposition and revenu stream. it just happens that the videos on this list A are the most extreme due to what we already know of human psychology and past TV studies. The algorithm will eventually automatically build a chain of recommendation of increasing extremeness, because that's what works better for the result it tries to maximise)
The sad thing is that this has been also proven to increase the feeling of insecure.
So, yes, initially the youtube algorithm will show up a variety of similarily themed video recommendations, some of which "must be a bit more radical or a bit less radical". But eventually some of these recommendation will prove more popular and youtube will learn to show them more. Due to how human psyche works, those more successful videos *will be* a little bit more frequently the more radical ones. And thus youtube learns to show the radical more a bit more often (without even having the notion of what "radical" is, only that they are successful). And again, sadly due to how human psyche works, it will have a negative impact on viewers.
It's hardly new to the internet :
it's been discussed for quite some time that TV channels, specially their News, tend to over-dramatize news and cast dramatic negative light on the world.
Basically, when creating emotion in viewer, their attention increases.
This works even more with negative emotions (fear, etc.).
This attracts more eyeballs to your channels,
giving you more opportunity to sell these attentive eyeballs to advertisers and thus increase your revenue stream.
In Europe, this is more prominent on private channels (mostly paid by ads) than on public channels (partially paid by taxes).
The thing is that, in practice, it has been proven to have an actual effect :
in Europe, watching TV and watching news even more so, has been linked to causing an increased feel of "insecurity, danger, etc."
This is despite the situation in Europe being much better and safer than before.
Criminality rate is globally decreasing, but TV reporting thereof is on the rise.
The neural-network AI used by youtube to process recommendation has simply rediscovered on its own the same results as what was already found on TV :
increase the emotional response of viewer by showing more extreme videos, you attract their attention and thus can sell more ads.
The AI doesn't even really have an actual concept of "emotional response" and "increased attention". The AI only notices that after recommending some video, revenue increase.
If video B is shown after video A : more retention, increased ads revenu. If then video C is shown after video B : again more.
AI the remembers to use the chain A -> B -> C, because that's what increases the parameter it is optimising for : eyeball time sellable to advertisers.
But because of what we know from what was studied on old school TV, that will eventually mean showing more and more extreme videos, because that's what has been proven to work on human brains. This AI has simply "independently rediscovered this fact".
----
For shit and giggles, let's gice a source that's on Youtube itself (and even parodies the usual style of conspiratorial videos). Sorry it's in French, but it has English CC.
Interestingly, The Concorde stopped flying in 2003, the very year Skype launched.
Funny, I simply used Skype as a metaphor. Didn't think about the timing.
But that had zip to do with businessmen/women's reason for not using it. Business travel has not decreased. Did you know that millennials are the most frequent business travelers?
I'm not saying that business travel is decreasing.
I'm saying that there's a very specific subtype of business travel ("This is a very important meeting, you need to go represent the company, now !!!") that require urgently transfering a business person as fast as possible on the premise.
That specific type of *urgent* business meeting has decreased, and thus a small barely significant use-case for concorde has disappear.
I'm not saying that business would save the concorde. I'm simply say that's about the single (rare) situation case where one might legitimately have an use of the ultra-sonic speed of concorde (but not making economical sense).
For everything else, it was basically a "showing off" project.
"2. Nuclear fusion will come next. It is just an engineering problem."
Funny how in the previous half-century we managed to solve every other "engineering problem", we weren't able to solve that one.
"It's just an engineering problem" == the science around it is basically understood. We don't need to discover or develop new science (as opposed to true anti-grav hover boards). But it is a ginormous project that will take decades just to build.
To take a space race metaphor : Once Sputnik is in orbit, it doesn't take a long strecht of imagination to think that humanity will also manage to put people on the Moon or built a giant international space station. But from the Sputnik point in time to the modern day's ISS, it took a ginormous amount to resources, money, time and engineering.
The 15 years time frame touted by the project is probably over-optimistic. But hey, we live in the "startup / kickstarter / etc." era. Where over promizing is an absolutely necessary first step to attract the attention of money sources.
We don't even have the Concorde anymore, how come that solved problem isn't even part of the landscape any more?
Solved problem/working solution DOES NOT mean "makes sense economically".
Concorde for most of its life was a vanity project that never manage to make sens economically and look a long-term sustainable project.
And that's even before the modern ultra-connected world removed even more of the few corner cases where Concorde might have been useful.
(Basically, Skype killed the need to have business man able to cross the world within hours - and these 3 business men where the only out of all passengers that actually had a practical use of Concord)
(Though nowadays, if it was still around, Concorde might successfully be marketed as an "exclusive" experience to a few ultra-rich 1%er with way too much money in their pockets).
In other words, all the /. who currently heavily criticize the economical sustainability of Tesla/Uber/etc.
would probably thing that those are still perfectly long-term stable plans compared to Concord.
Fusion, like space colonies, is a fantasy for weak-minded tech nerds looking for a religious experience to come from technology.
Space colonies are technically feasible and limited to engineering. But such a vast problem requiring so many resources compared to potential use now, that nobody is ever going to thow the necessary money at it. You would need dozens of government putting all their resources together for a century or two to even hope boot strapping it. (Or the US divert a few percent of your war budget :-P )
It's basically a project on the same scale as the colonization of the world by the big maritime empires, except back then the empire were having high hopes to gain riches, whereas there aren't much economic incentives to invest into space colonization now. Vague hopes of mining helium 3 on the moon do not sound a credible enough excuse to lock major part of the budget of dozens of government over the next couple of centuries.
And scientific discoveries aren't a good enough incentive. It's currently way cheaper to send probes and robot to explore instead of the costly and complex project of putting humans in person there.
Sorry, Bill. You won't see commercial fusion power, asteroid mining, or space colonies.
Well it's not clear when fusion will make sense economically - might take 40 years, might take a century - or might end up not making sense at all.
But the current big project like ITER, are currently done for the sake of science, as well as vanity project to show off engineering capabilites and try to learn useful stuff out of it.
We might manage to get something out of ITER - at least learn a lot - though it might take a lot more resource and time than initially planned.
TFA's startup ? well... I wouldn't hold my breath.
Anyone know how many apps out there actually use SGX?
On Linux (and other open source OS, i.e.: the only distributions where code for nearly all the software is available and can be recompiled with a retpoline-enabled compiler, such as the recent GCC 7.3.1, and thus the only environment where there's any hope for spectre counter measures to be actually deployed):
probably close do zero, anyway.
Not really. Not if both end stations have electric cars for rent that only need a battery range of 40 to 50 miles.
Which is actually a real-world situation in switzerland, with the biggest car-sharing cooperative (Mobility) also having cars available at trains stations, including electric cars in bigger cities (Renault Zoe - currently equipped with the smaller 125km range battery, progressively getting upgraded to the bigger 250km).
This is currently successful commercially.
no need for fast charging infrastructure either. Rail plus electric cars is a beautiful combination
Actually, due to how train work, you happen to have a fast charging solution available almost for free at the train station.
It is a translation layer, like WINE.
Cygwin is the WINE-like translation layer.
WSL is about the NT kernel able to speak Linux API directly.
(NT kernel is weird in that it can support several different API - used to also directly run OS/2 software without translations).
Clearly I/O is the issue, since compiling is very I/O intensive. {...} you will notice that the WSL is almost 100x slower than the bare metal version of Linux when compiling.
Because it doesn't matter if your kernel is able to directly serve IO request natively instead of needing translation...
When those IO request are served by the usually terribly sucky NTFS code.
The main IO bottle neck (Windows' own awefull performance on anything filesystem related) is still there, no matter how many (insignificant) abstraction layers you removed above.
The major performance issues that remain are with I/O.
Because even if there's no translation happenning, you're still bound to the sucky NT Kernel filesystem drivers.
Otherwise it's actually pretty good, in some cases equal to or even slightly better than bare-metal Ubuntu performance.
Keep in mind that this has mostly to do with the way multi-processing is handled :
- Windows suck at multiple processes, because creating a new context ( fork() ) is a horribly inefficient process. (Whereas on Linux, it's almost a free action thanks to CoW facilities in the virtual-memory subsystem).
- As such when running a unix software using a software translation layer (like Cygwin), multi processing will suck.
- That's why multi-threading is popular in Windows world : there's no context separation, everything is done in the came context.
- The NT kernel introduced a new concept called pico-thread which are much more light-weight than regular Windows process to setup. These aren't available in Windows, but gives a way to the NT kernel to provide extremely light-weight multi-processing to Linux ELFs.
- Multi-processing works decently well on WSL (unlike Windows native apps, or Unix apps via Cygwin).
- But if you read the technical blogs at microsoft, you'll release that the managed to achieve pico-threads by throwing away some of the context isolation of actual multiple process.
(There's a reason while picothreads aren't available for production Windows software)
- So basically in purely multiprocessing/multithreading benchmarks (e.g.: thinks running on OpenMP) WSL can even slightly beat actual real linux, because Microsoft threw a lot of safety and security out of the window. (It's great for testing software, but do not ever contemplate using WSL in production. It's only to test software before deploying on real Linux).
- When benchmarks are mostly CPU oriented (e.g.: most of the media compression tests) - most of the CPU cycles are spent running the instruction to process the data, and they are the same no matter what OS they run on (i.e.: a cygwin compiled software would run just as fast, provided it was compiled with a similar version of GCC the optimize code the same way).
- Whenever a benchmark hits any other part of the NT kernel (example: file IO) the performance just completely collapses. It doesn't matter that there's no translation going on and that the NT kernel is directly service IO request it self, when that IO code just plain sucks. A deep overhaul of the NTFS code would be the only hope for WSL to suck a tiny bit less.
(3D API in theory could be an area where performance degradation won't be as significant : some manufacturer like Intel produce better Windows drivers than Linux, and other like Nvidia re-use basically the same drivers.
But OpenGL is among the long list of Linux APIs that WSL is not supporting in its (very limited) subset).
Nope.
Wine is a translation layer, it's a piece of user land software that maps incoming Windows API calls from native .EXE to the closest approximation achievable under Unix/POSIX systems (quite a lot is achievable actually).
WSL is NT kernel having multiple personnalities. In addition to be able to directly serve Windows API to .EXE, the NT kernel itself also able to serve (an extremely small subset of) Linux API to ELFs, including stuff that doesn't exist in Windows API and could not be done by a userland translation layer such as Cygwin (Windows API sucks at multi-processing, but the NT Kernel has recently gained pico-threads which makes it possible to support decent multi-processing for Linux ELFs, despite such a thing not existing in the regular linux world). (On the other hand you're still limited to the NT kernel's horrible IO)
(A long time ago the NT kernel used to do the same for OS/2 applications - run them directly without any translation in the middle).
And then cue in companies complaining how they now need much more CPU in their servers, because instead of blindly serving files in the clear, they need to secure traffic with HTTPS and check authorization to send data.
And complaining how it costs more bandwidth because every single piece of data must be sent again over internet and can't be cached due to access control.
And how some people will be capturing and storing the streams locally any-way.
Here is where the war is currently. Add in phones and tablets and other walled off closed ecosystems and you can see this is a bad road where freedom doesn't exist.
If data is kept locked on the corporation's cloud, do you even need the phone and tablet ?
You can just keep the software as web app, and have the users keep paying a recurring fee if they want to have the privilege of keeping to work with the data they left locked. (see Software as a Service, "Microsoft Office Online", etc.)
You invite someone over for dinner and they steal your cutlery and napkins.... what do you do?
The weird thing with digital media and where this metaphor breaks down, is that even if that someone stole your silver cutlery and silk napkins, next time you open the drawers of your kitchen, they'll be magically full again, and you can still invite someone else on your table with proper silverware.
The closest thing would be inviting someone for a private view of a master's painting in your living room. But the guy take a polaroid, so he can look at it whenever he wants (shitty quality of the analog compared to what digital media replication allows aside).
The main problem is that we still apply an absolutely out of date method to pay artists for creations - making the duplication the paying step, even if nowadays it's actually the simplest and cheapest step of the process.
The industry solution is to try to invent some magical snake oil that suddenly makes Polaroid not work in the presence of work of art.. Which is cryptographically impossible to achieve when the usual "Alice" and "Eve" are actually the same person.
The actually potentially successful solution would be trying to find better ways to pay artists to keep creating, while not taxing the simplest part. Patronage and crowdfunding seem to be potential parts of the solution.
Will probably use the "LE" (Low Energy) Bluetooth protocol (it's also under the Bluetooth umbrella name, but has nothing to do with classical Bluetooth, it'ss more some kind of ZigBee-like low power protocol that got phagocyted by Bluetooth).
Which is completely stupid in this context(*), but helps Apple add a bullet point on the marketing speech ("Low Power !") and helps them use a protocol (LEA) which isn't widely used by anyone else.
---
(*): Audio is continuous streaming which is not something LE excels at.
LE is good for quick short burst message that only happen every now and then (like a health device giving a status : think heart beat monitors).
We already had this discussion back when Apple was spinning some marketing about having made LEA available to auditory aid equipment and thus being better than Android for people with hearing deffects.
But 30-40 years ago, the DCF77 standard started to become very popular.
Instead of having clock that still need to be set after a power outage or needs to be set after each daylight saving time change, you have a clock than can auto-adjust it-self hourly over the radio from an atomic clock in Germany, that still works across most Europe (i.e. within the long range radio's reception), is only a few seconds off (the ping time of the radio transmission and electronics processing), and has a notion of DST (so no need to manually move 1 hour forward or backward), and can even work on battery without anything plugged in (it's radio).
And it's German, so nobody will complain that it's tremendously over-engineered as a solution.
At some point in time in the past, TV stations radioemitter could use DCF77 as a source for the clock they broadcast over Teletext/Videotext pages.
But the cheaper "quartz china clocks" leave out the pricey part to save a few cents per unit and the time spent tuning it.
This *Europe* you're talking about. With a tendency to love over-engineered solutions (specially the german part of it).
Since the late 70s we we actually have a standard for automatically seting clocks : DCF77, and since the 80s radio-controlled clocks have been raising in popularity (I still have one from early 90s).
This things will automatically self-adjust every hour if they can manage to catch the signal. And the whole system is at worse a few seconds off.
No end user needing to fumble with some adjustment screw.
That's basically what they tried to do.
But they fucked up because the "free shareware episode" here already contains the data for the extra "commercial episodes" and only a small switch gets flipped to enable continuing.
(Whereas, you got extra floppies with Commander Keen episode 2 and 3, or Wolfenstein Episodes 2 to 6, once you paid after testing the free shareware episode 1).
(Though I've read somewhere that Id *DID* make the same blunder with the shareware CD of Quake1. There was only a simple registration procedure that was supposed to unlock the full commercial game with extra episodes out of the shareware CD, the same way here the full Yakuza6 can apparently be unlocked out of free 36GB demo).