1 |
On Mon, 04 Feb 2019 15:13:36 +0100 |
2 |
Michał Górny <mgorny@g.o> wrote: |
3 |
|
4 |
> On Mon, 2019-02-04 at 15:04 +0100, Alexis Ballier wrote: |
5 |
> > On Mon, 04 Feb 2019 14:54:40 +0100 |
6 |
> > Michał Górny <mgorny@g.o> wrote: |
7 |
> > |
8 |
> > > On Mon, 2019-02-04 at 14:48 +0100, Alexis Ballier wrote: |
9 |
> > > > On Mon, 04 Feb 2019 14:28:28 +0100 |
10 |
> > > > Michał Górny <mgorny@g.o> wrote: |
11 |
> > > > |
12 |
> > > > > On Mon, 2019-02-04 at 11:58 +0100, Alexis Ballier wrote: |
13 |
> > > > > > On Sun, 03 Feb 2019 20:28:49 +0100 |
14 |
> > > > > > Michał Górny <mgorny@g.o> wrote: |
15 |
> > > > > > |
16 |
> > > > > > > --- |
17 |
> > > > > > > What do you think? |
18 |
> > > > > > > |
19 |
> > > > > > |
20 |
> > > > > > What is the difference with sunrise ? |
21 |
> > > > > |
22 |
> > > > > The difference, as noted in the mail, is that it doesn't rely |
23 |
> > > > > on developers having time to review ebuilds. Therefore, it is |
24 |
> > > > > less likely to die because of developers lacking time to |
25 |
> > > > > review stuff. |
26 |
> > > > |
27 |
> > > > |
28 |
> > > > Then I fear you will see the same pitfalls, and it already |
29 |
> > > > started: I recall sunrise haters being very strongly against |
30 |
> > > > the idea because, TBH, our sandboxing mechanism isn't a real |
31 |
> > > > sandbox. It may have improved, but I doubt it's up to the point |
32 |
> > > > that we can safely run untrusted code there. |
33 |
> > > |
34 |
> > > Sandboxing has nothing to do with security, and trying to |
35 |
> > > 'improve' its security is a waste of time. What's the point of |
36 |
> > > preventing ebuilds from doing malicious things at build time if |
37 |
> > > they can install files that do malicious things afterwards? |
38 |
> > |
39 |
> > |
40 |
> > Because one may or may not run a malicious binary. You are more |
41 |
> > likely to install it. And even more likely to source the ebuild. |
42 |
> |
43 |
> 1. There are trivial ways to make you run something. Imagine an |
44 |
> ebuild installing into /etc/local.d. Or /etc/cron.d. |
45 |
|
46 |
|
47 |
Think of an automated QA check for those ebuilds, running some sanity |
48 |
checks and then uninstalling them. |
49 |
|
50 |
> 2. By design, postinst is run with full privileges. It is meant to |
51 |
> allow ebuilds to run stuff, as root. |
52 |
|
53 |
|
54 |
And that is precisely that kind of design that makes it hard or |
55 |
unrealistic to have unreviewed global repositories. |