Re: Bug#739657: gnuhealth-server: fails to install: gnuhealth-server.postinst: sudo: not found
On Mon, Feb 24, 2014 at 01:57:07PM +0100, Andreas (Debian) wrote:
> > Although for a package that requires more than 400Mb of dependencies
> > (Tryton takes in parts of LibreOffice, etc.) one extra dependency on a
> > package that is installed on 76% of machines (more details below)
> > shouldn't be a deal breaker (and that's in line with what you both
> > express in the next paragraph)
>
> I'm aware that if you simply count the installed bytes the small sudo
> package does close to no addition. I'm rather concerned about the
> principle to stick to the most simple way to approach a goal - if you
> can do it with a standard tool of coreutils you simply should do it this
> way.
That argument does not hold: coreutils will make things
*possible*, not *simple*. Or else we need to define
"simple" first.
> I think it is the other way around: Any code you install without really
> needing it might introduce some security whole. So simply don't do it.
Unless the code you install (sudo) is more secure than the
code that's already there (su) but doesn't get run as often
due to the other code being installed.
> Considering that GNUHealth is running in critical environments like on
> servers in hospitals you just want to minimise the intrusion vectors.
> Not installing a not really needed package that might give you root
> access is IMHO a vital advantage.
It is good practice for critical environments to not let users
log in as/become root but rather *do* carefully constrained
things as root -- and thusly, use sudo.
I'm just saying things can be looked at both ways.
Karsten
--
GPG key ID E4071346 @ gpg-keyserver.de
E167 67FD A291 2BEA 73BD 4537 78B9 A9F9 E407 1346
Reply to: