Quotes of the week
Posted Mar 24, 2012 14:03 UTC (Sat) by khim
In reply to: Quotes of the week
Parent article: Quotes of the week
The Android approach of giving each application its own UID/GID sounds reasonable here.
Yes, it looks reasonable on first glance, but, unfortunately, in practice it's not enough. Linux kernel is too large and too bug-ridden. It's developers have not actually considered case where people will run potentially-hostile-code locally. Simple UID/GID separation is not robust enough. This means you need additional sandbox. There are many choices: seccomp or ptrace can be used to create adequate sandbox, NaCl offers even more robust one.
This could be as straightforward as requiring a particular version of KDE's and/or GNOME's basic libraries to be available (after all, GNOME stuff can run under KDE and vice-versa) and getting the KDE and GNOME people to provide ABI backward compatibility with that version.
Yeah, this should be possible, too. Will need some amount of upstream support, but should be doable.
If you'll think about it then you'll understand that app stores are all the rage today because they offer two things simultaneously:
1. Distributions package is created by application developer itself, and it's easy to add it to the app store (review process is [relatively] simple for Apple's AppStore and post-moderated for Google Play Store).
2. User can install random programs from the app store [relatively] safely (Apple review process promises to weed out malware while Google does the same using combination of automated process and post-moderation).
Traditional distribution model offer only #1 while distribution repos offer only #2.
It should be possible to create something like this but then you need to jumpstart it somehow (people will only use it if it'll offer something not available from regular Linux repository and developers will only bother with it when there will be enough users).
That's why I recommend not start from scratch but to reuse (and later extend) format of one of the existing stores.
P.S. Also note this magic combination existed for 10 years before Apple and Google created their offers. This is how webapps have come to prominence! The only problem with webapps was creation of them: yes, it's easy to deploy the webapp and it's easy to convince user to try one, but it's not easy to write it in the first place! That's why webapps are all the rage on desktop but “meh” on mobile. They are not needed there! And now we have this and this. What does it mean for the future of Linux desktop? Nothing good: pendulum is almost ready to go back and reverse the trend of previous decade WRT webapps. The trend which made Linux desktop somewhat viable in first place (after all if the only program you ever run is the webbrowser then you can as well use free OS to host it and save money). Think about it…
to post comments)