Libertine: Allowing X11 Debian Packages To Run On The Next-Gen Ubuntu Desktop

[duby229]

So have fun fighting those tiny ms in the world of Care Bears ( https://en.wikipedia.org/wiki/Care_Bears )

Real developpers and companies were waiting snap (and others) to bring new products, games and applications to Linux in a simple and effective way. And this is much more important IMHO.

I fully agree with you that some kind of container technology will be the future, but snaps just don't do it right. That's all I'm trying to say. You may think performance isn't important, but it really is.

[bkor]

Ubuntu already has a native browser that does not use X11, it works fine on my Unity 8 desktop, phone, and tablet. You also have the choice of installing one of Google's Chromish browsers, or Firefox, or Opera, or whatever tool you want to do you job even if it means doing it through legacy support. If you want, you could even install a remote desktop application and run your Internet Explorer on a Windows XP system if it's really important to you that the browser be a part of the OS. Or no browser at all if your IoT toaster doesn't need to be used to sell your eyes to advertisers.

Initially you mentioned that the browser is not part of the operating system and it is separate. When I respond you tell me there's a native browser. What is it?

[Passso]

I think I get it, you think a millisecond is tiny, but really it's a vast amount of time. Think of it in terms of duty cycles. What you consider negligible is a 99.9% idle duty cycle.

So have fun fighting those tiny ms in the world of Care Bears ( https://en.wikipedia.org/wiki/Care_Bears )

Real developpers and companies were waiting snap (and others) to bring new products, games and applications to Linux in a simple and effective way. And this is much more important IMHO.

[ssokolow]

Take GTK+ out of that list, they explicitely decided to drop windows and osx support, as they are in their own words only concerned about the Gnome Desktop.

Yet another reason I chose to migrate my development efforts to Qt. (I'm not the only one who is unimpressed by the planned solution to the GTK+ 3.x API stability mess.)

[Mystro256]

First 150MB more on the installation phase is not important, as long as the daily use and launchtime is not impacted, but it seems we agree on that.
Then, 150MB is the total library size included in the snap, not the actual loaded size every time you run the program... and be it loaded from /usr/lib or /var/... changes nothing.

I really do not see your point. But maybe a benchmark could prove who is right more than words, and I am pretty sure the difference would be negligible

Windows turned into the mess it is today because there was a lot of "negligible" additions. These things can add up, and not having a common runtime available for snaps seems like a horrible design oversight.

[Mystro256]

And... please tell me a real life example where one needs to run a program multiple times?
LibreOffice for example open all docs in one running exe.

It's the libraries, not the main binary. LibreOffice for example links to other libraries, it's not a self encompassing binary.

For example, you open LibreOffice then Gedit. As snaps, there is no runtime, so it may open two instances of the GTK library in memory. With a runtime, such as the one in flatpack, it would open libreoffice and gtk, then open gedit and share the already loaded gtk instance in memory.

Of course this is a simplistic way of looking as it, especially because there's plenty of other libraries that can be used in common between two running flatpack apps, such as cairo, glib, libinput, etc.

[duby229]

First 150MB more on the installation phase is not important, as long as the daily use and launchtime is not impacted, but it seems we agree on that.
Then, 150MB is the total library size included in the snap, not the actual loaded size every time you run the program... and be it loaded from /usr/lib or /var/... changes nothing.

I really do not see your point. But maybe a benchmark could prove who is right more than words, and I am pretty sure the difference would be negligible

I think I get it, you think a millisecond is tiny, but really it's a vast amount of time. Think of it in terms of duty cycles. What you consider negligible is a 99.9% idle duty cycle.

[Passso]

Because disks have huge latencies? It's not magic, if code is executing, then the executable had to be read from disk at some point in the past. And that point is where the bottleneck will be. It may sound small to you, but many small latencies add up.

First 150MB more on the installation phase is not important, as long as the daily use and launchtime is not impacted, but it seems we agree on that.
Then, 150MB is the total library size included in the snap, not the actual loaded size every time you run the program... and be it loaded from /usr/lib or /var/... changes nothing.

I really do not see your point. But maybe a benchmark could prove who is right more than words, and I am pretty sure the difference would be negligible

[duby229]

150MB difference is the snap difference, not the actual memory difference you get when you run the program... As libraries are loaded in both case where will the performance issue comes from?

Because disks have huge latencies? It's not magic, if code is executing, then the executable had to be read from disk at some point in the past. And that point is where the bottleneck will be. It may sound small to you, but many small latencies add up.

EDIT: So lets say you are a snap maintainer and app-A is packaged with lib-1, also app-B is packaged with lib-1. That's two instances of lib-1 that needs to get loaded. Every single instance beyond the first one is a disk bottleneck. So the the way Linux caches files is by location, so if exactly the same file is loaded twice from different locations, the disk cache can't tell that and it gets loaded from disk twice.

[duby229]

And... please tell me a real life example where one needs to run a program multiple times?
LibreOffice for example open all docs in one running exe.

That's not what he was saying. In fact it's the first run, even if it's the only run, that matters most.