Maybe we'll go back, but I kind of doubt it.
Maybe we'll go back, but I kind of doubt it.
I'd rather a native app speak the network than my network-speaking browser try to be a native app.
wait.
right.
I see the problem.
Qt works fine for me!
Building native versions on the developing end is a foolish idea not suitable for programs in general, and it is time we abandoned the approach. It made sense at the time, but much time has passed since then. Some software might benefit from the increased obfuscation, but other than that it's just shit.
I'm curious how Microsoft killed Java. I assume you mean Java on the desktop?
And by "library" maybe I meant "runtime"? I mean you just have to send the HTML and JS to the client, and everything else for running the app is already there. I guess now, you could send a Python file written to use QT and it would work equivalently on many platforms.
I was thinking more desktop (non-web) apps, really.
An IR-esque approach is good, just not LLVM. Then you're back to something like the JVM.
[1] https://developer.chrome.com/native-client/reference/pnacl-b...
It's still easier than rewriting all of the development tools, libraries, etc in javascript and for the browser, and your app will be responsive and fast.
If you want to get fancy, and you're willing to sacrifice compatibility with less-cool platforms, JS is generally within 2x of native performance, which is plenty fast enough to feel "responsive". http://asmjs.org/faq.html
Just like in the 90s...
Web developers might think what they get to do is fun, but as a user the idea that software I use is going to change overnight without my permission is crazy. I shy away from web software. (Also, all of it tends to run far too slow on my system anyway)
Imagine if, say, we had a docker:// URL scheme, where you could click a link and it'd stream you a container image that would then start an ephemeral instance of itself with a linked persistent data container, taking the rest of the URL as its command line.
Tight security can be solved by some sort of sandboxing that takes a performance hit and a web of trust. Computers that ran the program can confirm that it is not malicious. Then an app can be run without a performance hit. This is the model of Apple's app store more or less. Except it's not a web of trust but a centralized authority and there is no way to run untrusted code.
And it makes a lot of sense. Most apps that are worth using will be used often, no reason they should take a constant performance hit after trust has been established.
Mobile devices are designed to host a new generation of apps. The Web is great for expedient universal access, but it won't be a universal multi implementation runtime that works better than platform native runtimes. The Web is a pre-touch legacy platform as much as Windows. Take any tablet and try using it as a touch Web device without using native apps to get a feeling for the scope of the problem.