| |||||||||||||
Free is too expensive (Economist)Free is too expensive (Economist)Posted Mar 31, 2012 9:28 UTC (Sat) by khim (subscriber, #9252)In reply to: Free is too expensive (Economist) by anselm Parent article: Free is too expensive (Economist)
After all, the Google app store knows what sort of phone you have and can tell you beforehand whether an app will run on it It's the other way around. Application includes minimum version of API in it's manifest file. Store only blacklists applications by the third-party requests (for example carriers like to blacklist tethering apps), it does not proactively track the compatibility story. there's no reason why a similar approach could not work for Linux There is: Linux desktop has no stable ABI thus if application works in distro M version N it does not mean it'll work in distro M version N+1. And forget about distro MM! Google's app store works because hardware vendors certify their variants of OS, not because Google magically makes apps compatible with some random crap. It would be up to the app developer to provide the appropriate packages. Does not work this way. Sure, some packages will provide updated versions to support different form-factors (for example some games needed to redo the controls when Android phones lost trackball), but it's mostly up to the OS vendors. Typical app developer releases app bundle exactly once. There would in any case be no guarantee that any app was available for any distribution/architecture combination, much like Android apps today don't run on every single device out there. It guarantees that it'll run. It does not guarantee that it'll work. There are a difference. If some required hardware is just not available then you don't have much choice: it's up to the app to decide how to handle the degraded mode. In some cases you can use some new hardware (not available when app was released) to emulate old APIs (witness ICS phones without hardware buttons and sensors: they emulate them), but usually user just looks on the program and decides if he wants reduced functionality or not. App developer is not forced to do anything in any case. If we ever manage to agree on a reasonable standardised platform for third-party apps then so much the better, but it's not essential for the concept to work. Yes, it is. Without stable ABI the whole concept crumbles. Developers don't want to track OS releases. They may decide to redo some flagman applications to make them better integrated with new OS - and even then it's not guaranteed. Most applications will never be updated. That's why Intel is fixing this problem WRT Google Play Store, for example. Not Google and not app developers.
(Log in to post comments)
Free is too expensive (Economist) Posted Mar 31, 2012 12:24 UTC (Sat) by anselm (subscriber, #2796) [Link] It's the other way around. Application includes minimum version of API in it's manifest file. Store only blacklists applications by the third-party requests (for example carriers like to blacklist tethering apps), it does not proactively track the compatibility story. So what? You tell the app store that your machine is running Debian GNU/Linux 6.0 on an amd64 architecture. You pick an app and click on the »Download« icon, and the app store sends you a package for the app that will run on Debian GNU/Linux 6.0 on amd64. Everyone is happy. It may of course be the case that the app store doesn't have the app packaged for Debian GNU/Linux 6.0 on amd64. In that case you won't get a »Download« icon. You may get a »Ask the app developer to provide a package for my system« icon. This will help app developers to figure out which distributions to support. The app store could also publish statistics on which distributions have the largest number of supported apps available (possibly by topic such as »games« or »web development tools«), which may help people pick what distribution to use in the first place. There is: Linux desktop has no stable ABI thus if application works in distro M version N it does not mean it'll work in distro M version N+1. A popular app store may create pressure on distributors and upstream projects (such as KDE or GNOME) to get their act together and be more careful about backwards compatibility. Backwards compatibility isn't really a problem with Linux as such or with distributions, it is mostly a problem of upstream developer negligence or incompetence. Also, if somebody at some point does define a cross-distro ABI that actually does provide desktop apps with what they require, that could be an alternative to targeting individual distros. If an app targets »LSB-NG 1.1«, that package would automatically be made available to all distros that are known to support that standard. App developer is not forced to do anything in any case. My experience with packaging stuff for Debian is that it is usually not a problem to move simple desktop apps forward from one version of Debian to the next, as far as the actual packaging is concerned. The ABI problem is potentially a different issue, but as long as you can get your stuff to compile and run, making it into a package for Debian N+1 isn't a big thing if you have already made it into a package for Debian N. The app store concept would make most sense if it came with tools that would assist app developers with actually providing packages for different popular distributions (again, the SUSE build service comes to mind as a possible model). The whole system could – and probably would – gradually move towards a more standardised platform. Doing it the other way round, by trying to provide the standardised platform first and hoping for people (and distributions) to adopt it, is probably going to be less successful. What will definitely not accomplish anything is sitting here and whining about how not only the present situation is terrible, but also how all the suggested approaches to ameliorate the present situation will never work. The Linux community does not work by revolution but by evolution, which is by its very nature difficult to direct. In the absence of a Steve Jobs figure who will lay down the law for everybody, the only thing we can do is offer something and hope that people will consider it a good idea and go along with it. Please outline how your app store concept is going to make Linux a viable option for current Windows or OS X users while at the same time making the existing Linux distributions obsolete and not driving off existing users who like things as they are. (Note that »Google will sort it all out for us« is not an acceptable answer this time around.)
Free is too expensive (Economist) Posted Mar 31, 2012 12:42 UTC (Sat) by Cato (subscriber, #7643) [Link]
A cross-distro app store with a single package format is needed, with some degree of guarantee that apps don't interfere with each other (maybe a sandboxed filesystem tree for the app's own files and settings, with only data files in a shared cross-app tree).
This package format would reference a stable cross-distro ABI version that changes only every 5-10 years.
Once you have this, app developers could target the app store and package format for ABI version N, and then just focus on developing their app's features.
Although Android threw out a lot of other Linux features and mandates Java, it shows that a more extreme version of this can be done. I'm not suggesting Android is the way, just that a Linux based system can be extremely popular with developers with the right approach.
Most Android users have no clue they are even using Linux, and that's exactly how it should be.
An alternative to a cross-distro ABI and app store would be that one distro, probably completely new, adopts this and becomes extremely popular (more like Android than Ubuntu).
Free is too expensive (Economist) Posted Apr 7, 2012 1:54 UTC (Sat) by khc (subscriber, #45209) [Link]
I wonder if it makes sense to argue about what's needed or not. Clearly, if a wider user base is a goal (and some would certainly argue that's not a goal or top priority), linux hasn't progress much on the desktop. So I fail to see how keep doing the same thing is going to make significant progress.
Some of us believe that an app store like model is needed, and some don't, and it's clear that at this point we can't prove it by arguing about it. What about those who believe that it has a chance group together and try it instead? I mean, it would definitely be a better use of our time than arguing about it on lwn.
Free is too expensive (Economist) Posted Mar 31, 2012 16:09 UTC (Sat) by khim (subscriber, #9252) [Link] Note that »Google will sort it all out for us« is not an acceptable answer this time around. You want details? I'm afraid I can not oblige at this point. And not even because I'm bound by NDA, but because the whole strategy is not yet finalized. The observation is simple: Google has two separate app stores and that's one too many. The question is: how to better merge them together. Different approaches are possible: you can invent some way to run Android applications in normal Linux distribution (perhaps even in browser?), or you can try to make sure Chrome Apps will be comparable to native Android apps on phone (right now they are quite limited), etc. IOW: it's not yet clear yet just how the desired thing will be created, but it's obvious what is needed. The obvious goal is to create the single app store for all supported platforms (not all apps will work everywhere, but some will be available on all platforms: Angry Birds on desktop are not all that different from Angry Birds on mobile so why do they exist in totally unconnected worlds? Please outline how your app store concept is going to make Linux a viable option for current Windows or OS X users while at the same time making the existing Linux distributions obsolete and not driving off existing users who like things as they are. You are asking bikeshed color when it not even known if bikeshed will even be involved at all. There are sooo few existing Linux users that it makes no sense to actually care too much about them. Windows and MacOS are significantly more important: both systems include built-in App Store (Windows store is in planning stages), both are not going away any time soon and both use Android-incompatible foundation (MacOS is closer, but Darwin is still quite different from Linux). Some people care about Linux users simply because they themselves are Linux users, but noone ever suggested that they represent more then a footnote in the whole story. Linux technologies, on the other hand, are quite important because they are used extensively in Android and ChromiumOS already. That means that, of course, solution (if and when it'll materialize) will use a lot of them - and that means that it should be easy to include Linux desktop in the plans. But it's hardly feasible if the infamous “Linux community” does not care and where all suggestions to participate in plans on early stages are indignantly rejected. Doing it the other way round, by trying to provide the standardised platform first and hoping for people (and distributions) to adopt it, is probably going to be less successful. Well, time will tell. As you can see from the above the whole idea at this point is to increase the audience which application developer can target with a single package, not to gradually move towards a more standardised platform. Perhaps someone else will create a different kind of app store with build farms and multidistribution support - but it'll be different story altogether. It'll be interesting to see how successful they'll be.
Free is too expensive (Economist) Posted Mar 31, 2012 16:24 UTC (Sat) by anselm (subscriber, #2796) [Link] This is a very long-winded way of saying »I haven't the faintest idea.«
Free is too expensive (Economist) Posted Mar 31, 2012 18:09 UTC (Sat) by Cyberax (✭ supporter ✭, #52523) [Link]
That's actually a fair answer to the question.
Nobody really knows what's going to happen with Linux desktop in near future. There's that sense of fin-de-siècle in the air - the current situation is unsustainable and Something Has To Happen.
But history also shows us that this 'something' can often be quite unexpected.
We'll see.
Free is too expensive (Economist) Posted Apr 12, 2012 16:55 UTC (Thu) by nye (guest, #51576) [Link]
>So what? You tell the app store that your machine is running Debian GNU/Linux 6.0 on an amd64 architecture.
Oh my god. I don't think you understand how completely outside the realm of feasibility this is for the non-technical user.
When tech support people bitch about users who only know that they're running 'Microsoft', and know that 'the blue e' is the internet, *they're not exaggerating*. That's not hyperbole. It's a literal description of a sizeable proportion of the user base.
I don't think it's even a significant minority, but actually the *majority* of users who don't know what version of Windows they're using - assuming they even know they're using Windows at all.
The proportion of people who know what processor architecture they're running has got to be far, far below 1%.
Free is too expensive (Economist) Posted Apr 12, 2012 17:14 UTC (Thu) by khim (subscriber, #9252) [Link] I don't think it's even a significant minority, but actually the *majority* of users who don't know what version of Windows they're using - assuming they even know they're using Windows at all. Well, usually they do know that - Windows is too hyped-up to miss. What they don't know is where Windows ends and other things begins. I've seen the cases when people bought new PC (with Windows obviously) and vainly tried to find Excel there (which they have not bought). The proportion of people who know what processor architecture they're running has got to be far, far below 1%. This is quite obviously not true because Linux occupies about 0.5% of the market share and most Linux users know that.
Free is too expensive (Economist) Posted Apr 12, 2012 20:16 UTC (Thu) by anselm (subscriber, #2796) [Link] Oh my god. I don't think you understand how completely outside the realm of feasibility this is for the non-technical user. The machine will presumably know (or be able to find out) about its own distribution and architecture, so it can tell the app store on the user's behalf. A method to do this is the first thing a prospective app-store proponent would want to standardise.
Free is too expensive (Economist) Posted Mar 31, 2012 15:15 UTC (Sat) by welinder (guest, #4699) [Link]
Aren't you precisely the same guy who elsewhere chants "out with
the old and in with the new" _all_ _the_ _time_? I.e., ABIs are unstable precisely because people like you don't see stability as very valueable.
That makes you look like a troll.
Free is too expensive (Economist) Posted Mar 31, 2012 17:25 UTC (Sat) by khim (subscriber, #9252) [Link] Aren't you precisely the same guy who elsewhere chants "out with the old and in with the new" _all_ _the_ _time_? Well, kinda. If the old interface is not stable then there are no benefits in keeping it around. If I can not write SysVInit file once and use it for all distributions then I don't see why do you want to keep it around. I.e., ABIs are unstable precisely because people like you don't see stability as very valueable. Nope. ABIs are unstable because noone tries to keep them stable. It's Ok to introduce “one final distruption” if it's needed to keep ABIs stable from that point on: witness GLibC 2.x or X11R6+. But if something was declared stable then you, of course, must keep it around. It does not mean you can not introduce new, exciting things! For example DirectX 10+ is radically different from DirectX 1-9. Everything is different: internals, drivers, API, etc. Of course DirectX 9 emulation is provided - but this is only feasible when old version had stable ABI to emulate! "out with the old and in with the new" happens regularly in iOS (think multitasking), Android (think GPU acceleration), Windows (think WF), and MacOS world (think Launchd). But it does not mean backward compatibility is not important! These are different (albeit related) issues. If you don't embrace and accept new realities then you'll eventually be left behind (think PalmOS), but if you embrace them by dropping support for old applications then it's also hopeless (again: think PalmOS and webOS).
Free is too expensive (Economist) Posted Apr 1, 2012 9:28 UTC (Sun) by wertigon (guest, #42963) [Link]
This could be fixed with the following model though;
1. App store receives the source code from the developer.
2. Signs a deal about this code may be modified to fit different distributions, but not released in public.
3. App store keeps track of OS updates. and takes out a reasonable fee.
4. ???
5. Profit!
Free is too expensive (Economist) Posted Apr 1, 2012 9:57 UTC (Sun) by khim (subscriber, #9252) [Link] 1. App store receives the source code from the developer. And this is where the story ends. It's easy to convince junk producers to give you their code. It's very hard to do the same with well-known (and thus desirable) programs. They perceive their code as one of the most precious things and emphatically don't want to send it anywhere. Do you know that Google does not have access to the Flash plugin sources, for example¹? Even if it has explicit agreement with Adobe and ships it with Chrome. Stallman may dislike it as much as he wants but proprietary software will not go away any time soon. Either your accept and accommodate the fact or you are destined for the obscurity. There are nothing wrong with serving niche needs (gNewSense is fine distro for some people), but if we are talking about mainstream then this approach will not fly. ──────────
Free is too expensive (Economist) Posted Apr 1, 2012 12:41 UTC (Sun) by wertigon (guest, #42963) [Link]
But the thing is, in this model you don't give away your code to the general public, you merely give the code to a maintainer that is specialized in porting software. It's no different than hiring a consultant to scrutinize security bugs in your code, except instead of looking at it they package it.
Such a model would fix most problems, but someone need to put up an app store and make it work with various distros.
Free is too expensive (Economist) Posted Apr 1, 2012 16:44 UTC (Sun) by khim (subscriber, #9252) [Link] But the thing is, in this model you don't give away your code to the general public, you merely give the code to a maintainer that is specialized in porting software. So what? You still send your “crown jewels” to some tiny company which probably does not have enough money in the bank to properly reimburse you in a case of the leak. Note that typically companies don't even have the source rights for all the components (witness mighty struggles when they want to open-source something) so even if you'll manage to placate internal legal dogs it's still will be mighty struggle to obtain sources for all the middleware used. Huge undertaking. And for what? For 1% of users? May be 2% if we are lucky? Are you joking? Such a model would fix most problems, but someone need to put up an app store and make it work with various distros. Sorry, but no. Such model may help with open sources packages (and thus it's probably worthwhile) and will include much derided “glorified bookmarks”, but most serious developers will just ignore it.
Free is too expensive (Economist) Posted Apr 2, 2012 9:15 UTC (Mon) by wertigon (guest, #42963) [Link]
But the thing is, it's still no different than hiring a consultant. It does not have to be a small company - Canonical, or Novell for instance, would do rather nicely. So yes, I obviously believe this model has a future, and it's a better one than staticly compiling everything always...
Free is too expensive (Economist) Posted Apr 2, 2012 9:41 UTC (Mon) by khim (subscriber, #9252) [Link] But the thing is, it's still no different than hiring a consultant. Consultant usually is given access to small piece of the code at time. Often it agrees to sign extensive and complicated agreements to see even that. And in any case that's for serious issue (security) which warrants serious efforts. 1% of market is just not enough to do something like this. Heck, 5-10% of the market (which MacOS owns) is not enough: companies which are contracted to port stuff to MacOS often receive some things only in binary form (that's why CrossOver for Redistribution is popular way to go in these cases)! IOW: consultant you are hiring plays by your rules and must convince you to do this or that. Access to the whole codebase is rare luxury which consultant rarely gets! Yet here you are starting from that premise… It does not have to be a small company - Canonical, or Novell for instance, would do rather nicely. So what? Google is larger than both of them and it still is not trusted enough without extensive agreement negotiated separately for each piece of software. This process just does not scale.
Free is too expensive (Economist) Posted Apr 2, 2012 15:28 UTC (Mon) by jedidiah (guest, #20319) [Link]
I don't even think it needs to be source code. It could just be the binaries. Then the "store operator" would be in charge of packaging. Some of the Loki installers have been redone for newer versions of Linux. That kind of approach could be done wholesale.
A universal frontend could be done for the old games by someone like Linux Game Publishing as a proof of concept. They could even throw in some old apps too.
Distribution support would be driven by demand of course.
Free is too expensive (Economist) Posted Apr 2, 2012 18:41 UTC (Mon) by khim (subscriber, #9252) [Link] This plan will fall apart because distributions remove old versions of libraries willy-nilly. Sure, if you'll first create some stable ABI and will keep the set of libraries for that ABI around then it'll work… and this is what this thread is all about.
|
|||||||||||||