That makes sense to be honest, I thought it was intentional but then seeing so many other maths errors in other DFW works led me to believe it might very well not be.
>The SDK is there to allow people to add whatever features they want. If there is something they can't add with what the SDK provides, then either it requires changes breaking component compatibility (which only I could do even if the source was open), or person trying to implement the feature is doing something seriously wrong (happens very often).
"Implementing that feature would break component compatibility" is not a valid reason not to release the source. If someone wants to modify the software to implement a feature they want even if it would break compatibility, that's their business.
>As for porting to different OSes, sourcecode release won't magically spawn people capable of doing that properly. Somehow no one has written fully functional foobar2000 clone yet.
The point of having it open source is that the possibility is there. Right now it's impossible. Someone has to go through the trouble of documenting all the features and then reimplementing them.
>Sourcecode loss argument is not really valid, I keep backups on multiple redundant devices. I'd be surprised if someone who spent as much time on programming as I have wouldn't know well enough how to handle this.
Two words: bus factor.
I see attempts to refute reasons to open source the code, but no reasons not to do it. If the reason is simply "I don't want to", that's perfectly fine, and it's all that needs to be said.
I think people like this strive for control. Their projects are like their little kingdoms where they have the last say. You might say that they can still retain such total control even in an open source project (OSS doesn't necessarily imply "democracy"), but there's still a possibility of a vim/neovim-like split. Bram was also quite opinionated, which led some developers to fork vim. Bram was very clearly quite unhappy about the split in the community, and keeping sources closed will prevent such a scenario.
I’m not so sure about that. As a GNU/Linux user, I don’t think Windows-land has anyone who can cause as much community division as Lennart Poettering did (does?).
> Somehow no one has written fully functional foobar2000 clone yet.
Deadbeef [0] may not be "fully functional" because it doesn't support foobar2000 plugins or some such silliness but it is close enough to play the music library I played under Windows with foobar2000.
All of the listed reasons are humbug, as someone in that threads points it out. The only real reason is that the author wants it so, and that's why it happens. No particular reason or supporting argument is stronger than this will alone.
- A genie hands the encoder the type of an iid source sequence (the function N here: [1])
- The encoder produces a representation of the source with fewer bits than what would be possible using a Shannon-efficient encoder that didn't know the type.
- You also need to know the type to decode.
The reduction in bits from Shannon's theorem is explained by the genie revealing this extra information that reduced the input sequence's entropy (also causing it to become non-iid).
The collection of letter-typical sequences of specific type is indeed often going to be smaller than (entropy-rate)^(blocklength), and that is what we are seeing here.
This is where the author's explanation starts to go astray:
> The Shannon limit uses Stirling's approximation which is an asymptotic approximation for factorials and as such the approximation is too large for small values and becomes more accurate as the factorial size (i.e. message length and symbol count) approaches infinity.
I would recommend the author read any information theory textbook, starting at the latest, at the part that covers Asymptotic Equipartition Properties. This is the crucial definition that will make the source coding theorem concrete, in my opinion.
The immoral part is how badly many elementary through high schools teach math.
As a math-adjacent (CS) professor, it was painful to watch my daughter's math experience at her first elementary school. Part of the reason she's not there any more, in fact.
The Wikipedia link on "lattice" points to the wrong article. "Lattice-based cryptography" usually means group-sense lattices, not order-sense lattices. Lattice groups are even directly involved in the rest of the article.
I got a Room and Board 65" Jasper in Cognac Leather right before the pandemic. I thought I was overpaying and it ended up sitting in shipping for 4 months because of lockdowns. I was predisposed to hate the thing but it's become one of my most reliable large purchases. Very solidly built. The leather has held up perfectly in front of bright windows.
This is the Spectral Theorem. It says (in the version for real matrices, which are what we have here) that any symmetric matrix has an orthonormal basis of eigenvectors, or equivalently that any symmetric matrix can be diagonalized via an orthogonal change of coordinates.
Fairly recent Hondas and Acura base infotainment systems have needless >1 second Bluetooth audio lag. I wish someone with the knowledge would do something similar for these.
Don't all cars have this? I've driven lots of different cars from all kinds of manufacturers and they all have Bluetooth lag, sucks if you want to watch YouTube just sitting in the car waiting for something.
I blame this on crappy drivers. A lot of modern Infotainment systems are running Linux of some variety, and the very easiest thing to do to get technology support is to grab the open source proof-of-concept Bluetooth driver, plug it in, and call it good.
As a firmware developer who has also done client-side Bluetooth development on Android, iOS, Windows, and Linux, I can tell you with 100% confidence that this just isn't the case. BlueZ is the best Bluetooth driver on any major platform, and it's not even close. The Windows one crashes constantly, and the iOS one has arbitrary nonstandard limitations and it tries to be clever by initiating pairing when you don't tell it to. BlueZ is the only one that actually implements the by-the-book Bluetooth spec, and does it without randomly exploding.
No, the reason Hondas and Acuras have it is specifically because of (if I understand it correctly) a bug in the car’s Bluetooth handshake that sends audio streams as data. Someone found this by doing a Wireshark capture connecting a laptop’s Bluetooth to the car.
I have little hope for a fix. It seems likely that its easier to just replace the entire controller that drives the audio and screen, and that sounds really really hard.
