Proprietary software in Linux

[ean5533]

Maybe it won't be perfect unification, but unification which should be good enough to satisfy some third party vendors. For example if Ubuntu, Fedora, OpenSuse will use systemd (it's on the good way already), pulse audio (this one is used by all of them) and Wayland (it seems all of them are interested) companies can focus on those three distros. I don't believe they'll accept some unified package format soon and this will have to be resolved somehow, but it shouldn't be hard. Other distros like Arch, Gentoo don't seem to be interested in commercial support too much, but it seems they won't be left behind - they evolve around current Linux's stack, so it's very easy to install pulse or systemd on them. In my opinion if some other distribution will be interested in good commercial support it will follow the UFO.

I think it's an excellent idea. If those three distributions could agree to a "standard libraries" spec that vendors can refer to, then closed source apps could bundle their own additional libs against a known base and everyone would be happy. Other distributions could either meet that standard lib spec, or users could be educated about how to make their install match up with that spec. That would eliminate the whole "moving target" problem that vendors run into so often.

So, who wants to go propose this idea to Mark, the Fedora board, and Novell?

[johnc]

I'm probably one of the morons that thinks package management for the user (on Ubuntu at least) is a step ahead of Windows. To me it's much easier, though I do think there are some changes that would be nice. But as it is now, if it's in the repo, or a PPA is available for the software, to me it's a million times easier to install and keep up to date than on Windows. It's one of the reasons I've decided to stick with Ubuntu.

Now developing and distributing software for a broader linux is probably a much more difficult manner, since you can't just come up with a PPA to drop your .debs in.

I do think that this area has improved so vastly that it's worth examining as a case study of what's been done right. I remember fiddling with linux for awhile in the past and the most recent time was just as RPMs were starting to come about. Getting even the simplest software was a pain in the doofus. You had to get the source, then find all the dependencies, then compile the dependencies, then find all their dependencies, ad nauseam. I mean I'm a software engineer with a comp. sci. degree who has been dinking around with software since the sixth grade and even I didn't have any patience for that crap. If I wanted an mp3 player I just wanted to get it, install it, and run it... not work on compiles for 40,000 hours. But I'm telling you, there were people back then that thought that's the way it should be.

But all in all, if there is a user base for commercial software vendors to target, they would find a way to drive their software to consumers.

[ean5533]

what is wrong with pulseaudio?

they dev pulse audio to fix thix problem. the only problem is old software do not support pulse audio.

The only time I tried to work with audio was about 3 years ago, back when PulseAudio was a buggy mess and was very poorly integrated with the few distros that used it. I haven't attempted it since then, though I've heard that the situation is much better than it used to be.

To be fair, I probably shouldn't be running my mouth about something I haven't tested in years.

[89c51]

I'm probably one of the morons that thinks package management for the user (on Ubuntu at least) is a step ahead of Windows. To me it's much easier, though I do think there are some changes that would be nice. But as it is now, if it's in the repo, or a PPA is available for the software, to me it's a million times easier to install and keep up to date than on Windows. It's one of the reasons I've decided to stick with Ubuntu.

Now developing and distributing software for a broader linux is probably a much more difficult manner, since you can't just come up with a PPA to drop your .debs in.

but the main -or bigger- problem is not how the user gets his packages. there are some problems there but not that big in general. the devs -be it proprietary or FOSS- have to do a lot of job to get thing running and give them to the user.

and that hurts the "ecosystem".

[johnc]

but the main -or bigger- problem is not how the user gets his packages. there are some problems there but not that big in general. the devs -be it proprietary or FOSS- have to do a lot of job to get thing running and give them to the user.

and that hurts the "ecosystem".

Yup. That sounds like a disaster.

I guess if I was distributing a Java app it'd be a walk in the park.

If I had a non-Java app I'm not sure what I would do. Maybe make a .deb and a PPA and wish everyone else good luck.

If I was selling a commercial app I would almost have to statically link in all of my libraries into a single binary or maybe package in some shared libraries... like XBMC does. Then lay down certain system requirements to run. Or, more than likely, just not distribute for linux at all. In the mobile market the complaint is that Android is too fractured... but it's nothing compared to the rest of linux.

[pingufunkybeat]

This discussion is confusing. On the one hand, it is claimed that Linux is just a hobbyist toy OS when it comes to the desktop, on the other hand, people are claiming that Wayland is the solution. In fact, Wayland is a hobby toy system, at least it still is.

The interesting thing about Linux graphics (and input, and everything else related to X) is the fact that much of the complex stuff that used to be in X has been moving into the kernel. This is THE reason why we are even talking about Wayland. People have been talking about X-killers for decades, but it was only after GEM/TTM, KMS, DRI2, evdev, and other technologies became reality, that such a thing even had any chance whatsoever.

Wayland is a thin display manager, that is still just a hobby project. I expect that if mainstream Linux some day moves away from X, it will be something that has learned from the Wayland project, rather than Wayland being "the big thing" that "everyone is backing". It's just a thin layer over all the DIFFICULT stuff that has taken place, or is taking place, such as moving device handling deeper into the kernel.

What will likely happen is what happened with Keith Packard's toy project, KDrive, which was a testbed for technologies which all ended up in the X System. This is why Keith's perspective is really interesting -- he's been there, he knows the system inside out, and if he says it should be ditched, than it carries a lot of weight. X has changed a LOT in the last 10 years, and has been cut down in many important ways, and what it does today, on modern systems running modern toolkits is not what it used to do back in the early 90s. But perhaps the solution is not to keep adding things to it and modifying it to fit the modern desktop concepts, but rather to make something simple that only does what it needs to.

I will miss network transparence, but some sort of VPN-like protocol which lets me run SINGLE (!!!) applications remotely and have me control them from a host system would probably fulfill my needs. Many of the remaining X technologies are not really needed today.

But what is happening is that other Unix systems, which lack the commercial backing and the market share of Linux (like BSDs), are losing out, because all the things that used to be handled by X have to be rewritten or ported. This is very unfortunate, IMHO, but probably inevitable.

[89c51]

@pinkufunkybeat

what will likely happen is that wayland will replace X and probably will be used outside the desktop. and by the time this happens there will be an oh so cool remote thing, other cool stuff and legacy app compatibility with X.

[elanthis]

I'm probably one of the morons that thinks package management for the user (on Ubuntu at least) is a step ahead of Windows. To me it's much easier, though I do think there are some changes that would be nice.

There's nothing moronic about that sentiment. Linux package management has a ton of advantages over Windows. Windows' approach has advantages over Linux.

The problem is not "Linux is not just like Windows," but that "Linux could have a superset of the Windows features and ease of use, but instead distros explicitly and intentionally handicap the system." A system like RPM or DPKG is entirely capable of being every last bit as easy to use as Windows' for regular users with a few technology updates and a ****load of metadata and user-experience design work. It's not the formats that need changing so much as its the metadata in the packages, the incompatible update services, and the difficulty of distributing widely compatible installers.

The religious wars over package managers is part of the reason why I believe that a parallel installer is needed. A tool which takes some distro-neutral installer format and can install them while registering the installations with the native package system (if one is available/supported). Combine that with a firm approach to using _platforms_ rather than individual packages for dependency management (you've got the "OpenGL Platform 3.2" rather than "libmesagl-8.2" or whatnot, or "GNOME 3.2" rather than a long list of individual packages) and this tool could work quite well.

In particular, automatic updates could be handled way better on Linux than on Windows. Windows has no general update service for applications, so you end up with 50 processes running to check for updates for each app you have. It's pretty annoying, and there's no good reason for that other than Microsoft not putting a lot of attention towards such things.

Combine the above with a registration service and an app and website for browsing apps registered there and you'll get a distributed, non-centrally-controlled app store (that is, the packages are distributed and under no single entity's control; the listing service probably would be, but would be easy to replace if it ever went evil) that works for all conforming Linux distros. That would be hot.

[pingufunkybeat]

Actually, I will tell you what will really happen:

By the time Wayland is anywhere near ready to be even considered as an install option on a major distro (there's lots of work to be done there, much of it not in Wayland, but in toolkits, apps and supporting libs), X will have all the equivalent functionality and will actually be faster, more optimised, and more stable.

At least this is what had happened every single time there was an X killer before. And there were several.

[ean5533]

Actually, I will tell you what will really happen:

By the time Wayland is anywhere near ready to be even considered as an install option on a major distro (there's lots of work to be done there, much of it not in Wayland, but in toolkits, apps and supporting libs), X will have all the equivalent functionality and will actually be faster, more optimised, and more stable.

At least this is what had happened every single time there was an X killer before. And there were several.

If that's true, the Wayland will still have (indirectly) accomplished its goal of making Linux a better place, so everyone still wins.