WebAssembly Ends Browser Preview With Initial API & Binary Format

[ldo17]

Many developers do not want to expose their code to users to many reason (IP, security, etc.).

Downloading code to the users’ machines and expecting to keep it secret from them is just ... words fail me.

[ldo17]

A webapp can pull off a Continuum-like effect pretty easily (good websites do this all the time, it's called "responsive design") without requiring bullshit OS-specific integrations like Continuum, or the limitations of current cross-platform development framewors (that don't support 100% all features of all platforms, more like a 60%).

The Web is not magic. This just means the browser becomes the “cross-platform development framework”. And instead of supporting 60% of platform features, it becomes more like 30%, because of the security issues, limitations of JavaScript/HTML etc.

[ldo17]

E.g. Google Sheets isn't that responsive if you have a project with a dozen of sheets and like ~20000 lines, 30 columns per sheet. We actually managed to corrupt the G Sheet so it ended up read-only during simple manual data collection, and we started from scratch with a imported backup few times with the same corrupted results. Sure, this can be fixed, it's a bug.

Never use a spreadsheet as a database.

[starshipeleven]

Other thing is, it's obvious that the perf on the web platform scales really bad compared to e.g. gnumeric.

Little nitpick, Gnumeric isn't a office suite.

And of course I was not talking of major data crunching usage like yours, but what most office people do (a few orders of magnitude less).

I'm a bit worried with the quality of realtime apps in a browser. Maybe having each tab in its own process helps, but so far my Firefox extensions prevent this.

It can be sure fixed in FF, that's an immature feature for now.

Afaik Chrome is multithreading since years (each tab and also plugins/extensions run as a separate processes, it breeds like a rabbit), you might want to try that out and see if it can handle what you want, to have an example of "mature" multithreaded browser.

You can also try a clean FF out in a VM or by force-enabling electrolysis. I dumped all extensions that were preventing it and still had to force enable it, but I don't use FF as a player anyway, I only wanted to avoid the bullshit-page-hanging-the-whole-browser issue.

You just can't expect that kind of extreme performance in a browser.

Don't confuse performance with proper process isolation (and prioritization).
To run a music player you don't need a quad-CPU xeon-phi computing node, you need the player running on its own process, so that it won't be affected by other processes hanging or dying or whatever.

[caligula]

Office suites are huge terrible monsters while Google Docs runs in a browser and is great, and so on.

I kind of doubt they actually work with larger data sets. E.g. Google Sheets isn't that responsive if you have a project with a dozen of sheets and like ~20000 lines, 30 columns per sheet. We actually managed to corrupt the G Sheet so it ended up read-only during simple manual data collection, and we started from scratch with a imported backup few times with the same corrupted results. Sure, this can be fixed, it's a bug.

Other thing is, it's obvious that the perf on the web platform scales really bad compared to e.g. gnumeric. I can run the same sheet on a Pentium 3 when using gnumeric. Modern $1500 Skylake i5 ultrabook with 16GB of DDR4 has issues with performance even when the internet provides offers a 100M+ fibre. For example, when opening a sheet, it can take 6 seconds before the equations in the sheet get re-evaluated. When scrolling the sheet, it can be damn slow and unresponsive. With gnumeric no issues. Don't get me even started with RAM usage.

Native apps, yes, but they are either not truly cross-platform (a bunch of different applications), or to be cross-platform they suffer lower quality usually.

I'm a bit worried with the quality of realtime apps in a browser. Maybe having each tab in its own process helps, but so far my Firefox extensions prevent this. I play some MIDI/SID (C64) songs with a JavaScript player and the playback halts several times each day due to some hangup in other tab's JS engine. Needless to say, the same overclocked i7 7700k plays the same song just fine without a single glitch in a native player when I'm simultaneously converting h.264 to h.265 with handbrake (16 threads) and compiling the kernel with make -j8. Sure the CPU is hot @ 90°C, but it just works and schedules timeslices just fine for the player. You just can't expect that kind of extreme performance in a browser.

[starshipeleven]

Case in point, Visual Studio Professional 2015 sucks my RAM dry whereas I can have 20 unique Visual Studio Code windows at any given time and not see much of a difference in memory utilisation.

For the sake of giving some context, VSP2015 is written in C++/.net/C# while Visual Studio Code is web technology running on its own web browser engine (blink, from chromium)

[unixfan2001]

Case in point, Visual Studio Professional 2015 sucks my RAM dry whereas I can have 20 unique Visual Studio Code windows at any given time and not see much of a difference in memory utilisation.

The latter also happens to be multi-platform, whereas the former is Windows only.

I couldn't care less about "native".

[ldo17]

Office suites are huge terrible monsters...

Really? Which one do you use? LibreOffice seems quite good, particularly since they got rid of all that Java code.

[starshipeleven]

I think yszolt has a valid objection - native apps are more efficient.

Native apps like what? Because I think we need to make some differences.
Most modern applications aren't really written in pure C with some assembler optimizations, but in some degenerate bullshit ultra-high-level language that does not really have high performance to begin with, under time pressure and so on. The goal of the "ultra-high-level language" is to sacrifice performance for the sake of reducing development cost.

Cases in point: Skype always sucked hard, even on windows or Android where it should not suck. Office suites are huge terrible monsters while Google Docs runs in a browser and is great, and so on.

But the problem is, the native apps 99% of the world is using right now work on Windows, OS X, iOS, and Android. If we shift to more native apps, that just makes it harder for people to move to open source operating systems or even from Windows to OS X or vice versa.

Native apps, yes, but they are either not truly cross-platform (a bunch of different applications), or to be cross-platform they suffer lower quality usually.

Running a true application in the browser as a webapplication without having to rewrite the logic in javascript is very interesting for this.
A webapp can pull off a Continuum-like effect pretty easily (good websites do this all the time, it's called "responsive design") without requiring bullshit OS-specific integrations like Continuum, or the limitations of current cross-platform development framewors (that don't support 100% all features of all platforms, more like a 60%).

Then as a very secondary and unintended effect it will also let people on Linux/BSDs use the same apps (assuming the app-browser interface is the same).

[starshipeleven]

It's not for running desktop software in the browser. It's for having similar performance to desktop/native code.

yeah, but if you have a web application that runs more or less the same as a desktop native application, Then there is no reason why you should make a desktop native application for stuff that does not require top performance (like heavy gaming or workstation applications).