Quotes of the week

Posted Mar 26, 2012 11:03 UTC (Mon) by anselm (subscriber, #2796) [Link]

For example on Windows or MacOS. GIMP has his registry, TeX has TeXLive and Chrome has it's store. They all exist quite independently and don't really affect each other.

It turns out that there are various pieces of software that need TeX installed, or some sort of web browser. The usual Linux distributions handle this easily. Smartphones and the iPad generally don't run TeX and come with a web browser preinstalled, so the issue doesn't really come up.

How, in your opinion, would an ecosystem based on an »app store« handle this? Having each such »app« include its own copy of TeX isn't feasible as TeX tends to be somewhat large (usually way larger than whatever »app« is using it). Asking users to make sure that they also install the TeX app is a step backwards from what distributions do today and makes it difficult to claim that the »app store« is really more convenient for users. So it seems you're back to explicit dependencies.

Distributions promised to do the impossible: to support applications which pick dependencies at random without thinking. And now they are chocking under the load. The Linux desktop world is crumbling because you can't depend on it: familiar applications can be changed without notice (see GNOME and KDE) or just disappear (things like the venerable XV or GnoCHM) unless user itself decides to play Atlas, too.

This is arguably not the distributions' fault. They are doing a great job packaging things like the Linux kernel or glibc where the upstream authors do care about backwards compatibility, and they usually do allow several (major) versions of a library to coexist, which ought to take care of the ABI evolution issue if the upstream library authors do their job properly. Blame the GNOME and KDE developers for apparent incompetence or misguidedness or the XV developers for finding more interesting things to do with their lives. You can really only fault the distributions for not releasing often enough (if you insist on always having the newest stuff on the day it comes out) and not wanting to distribute stuff that they aren't entitled to distribute.

But it works better then the distributions model and as long as Linux desktop will reject it it'll have the same 1% of the market.

The »app store« model works for iOS because it is the only model there is (on Android you can install stuff from elsewhere but it is enough of a hassle for most people not to bother). Apple (and to a lesser extent, Google) don't use this because they want to make life easy for everyone but because they want to keep control over what is available and also take their cut.

As far as Linux is concerned, there is nothing wrong with the distributions model other than that there could be more ABI compatibility for the benefit of third-party app developers. LSB was supposed to sort that out but as we have seen was mostly a paper tiger.

At the risk of sounding like a broken record, the reason why an app-store model works for iOS but will probably not fly for Linux is that, unlike Apple at the time, the Linux community has no Steve Jobs figure. In the Linux community, any app-store model will need to succeed on its merits, not because some autocrat decides that an app store is how software is going to be distributed. You may argue until you are blue in the face that Linux distributions don't actually work but the fact of the matter is that, for all practical purposes, all existing Linux users run one distribution or another. These distributions set a bar that is uncomfortably high to jump over if you're an app-store-based model – because whatever their shortcomings the distributions are pretty damn good at what they are doing, namely putting useful software in the hands of their users. So the »app store« will have to offer something that distributions don't, and that will have to be (as I wrote earlier) convenient installation of uninteresting, or very new, or non-FLOSS software, i.e., stuff that distributions won't touch, and that out of the 1% of Linux users, probably only a small percentage is really interested in to begin with.

Your argument is that people who are not using Linux now will be enticed to switch to Linux based on such an app store, and it is an argument that I don't buy. I'm all in favour of more standardisation, and I think it would be great in many ways if there was an obvious »Linux« that new users could install to get started rather than the current mess of half a dozen reasonable choices, all with their different strengths and weaknesses. But I have also been around the Linux community long enough to be fairly sure that this isn't going to happen anytime soon. Linux could probably do with more of a market share on the desktop, but claiming that an »app store« for Linux will make all the difference there is cargo cult thinking – if you just flatten the grass enough and put on runway lights then the airplanes full of goodies will come to you.