Zero Install: An executable critique of native package systems (Linux.com)
Posted Mar 1, 2007 12:41 UTC (Thu) by
Tom2 (guest, #43780)
In reply to:
Zero Install: An executable critique of native package systems (Linux.com) by vonbrand
Parent article:
Zero Install: An executable critique of native package systems (Linux.com)
It doesn't 'allow' users to install to $HOME. Users can already install to HOME (see Autopackage, setup.sh, etc). What it does is lets users install in a controlled way (with GPG checks, regular security updates, etc) and efficiently (downloads shared between mutually untrusting users).
"If the application so installed has security problems, the system as a whole gets to be vulnerable."
Depends what you're comparing. If you're a sysadmin on a commercial system, your view of the outcomes might be:
1. User installs malware to HOME. Given an additional vulnerability in your core system, it might get root. BAD.
2. User fails to install software. System is OK. GOOD.
Even more secure would be hiding the power cable for the machine. Assuming the user isn't just messing about, and actually needs the program, the options are:
1. User installs malware to HOME. Hopefully contained in that user's account. GOOD.
2. User installs malware as root. Whole system compromised. BAD.
"If the program is worthwhile, surely some volunteer will show up for managing as an (extra)official package for her favorite distribution. If they don't show up, well..."
Just because it's possible to do something inefficiently it doesn't mean we should. We don't need combine harvesters, because there are enough people in the world to collect all the crops by hand. But perhaps it's better to have them spend their time doing something more useful?
(
Log in to post comments)