1 |
Hi guys, |
2 |
|
3 |
In the "Gentoo Hardened SELinux Development Policy" [1] we have a section |
4 |
requiring development to use the 'gentoo_' prefix. The reason for that was |
5 |
to ensure no collisions occur when a patch is added upstream. |
6 |
|
7 |
[1] http://www.gentoo.org/proj/en/hardened/selinux-policy.xml |
8 |
|
9 |
However, with the release of 20110726 and other changes, I'm pondering about |
10 |
removing this section from the guideline, and here is why... |
11 |
|
12 |
First of all, "safe migration" is not possible. We had around 40-something |
13 |
patches applied to 20101213 and less than one third could still be applied |
14 |
to 20110726. Not because the patch was included, but because the structure |
15 |
of the code had changed. All other patches needed to be made manual anyhow. |
16 |
|
17 |
Using gentoo_ prefix or not wouldn't make a difference here. |
18 |
|
19 |
Second, if a collision occurs, we would either get a failed patch (which we |
20 |
can then safely drop from our patch bundle) or a duplicate definition (which |
21 |
we will notice during builds, after which we can update our patches). |
22 |
|
23 |
Using gentoo_ prefix or not wouldn't make a difference here. |
24 |
|
25 |
Third, we are pushing many of our changes upstream. However, as long as we |
26 |
use different naming conventions, then the patches cannot easily be pushed. |
27 |
Currently, I'm manually typing over most patches that include gentoo_ |
28 |
prefixes into a reference policy checkout for submitting upstream, which is |
29 |
*very* time consuming. |
30 |
|
31 |
Using gentoo_ prefix is a time hogger. Using upstream naming convention |
32 |
would be much leaner. |
33 |
|
34 |
Fourth, supporting tools that help SELinux developers for a proper coding |
35 |
style as well as other documents and guidelines are often based on the |
36 |
naming convention. By using a gentoo_ prefix, these tools give warnings (and |
37 |
the documents are less valid). If we need anything at all, a suffix would be |
38 |
much more flexible. |
39 |
|
40 |
Using gentoo_ prefix here is causing development efforts to become more |
41 |
difficult. |
42 |
|
43 |
|
44 |
I'd rather use the gentoo_ prefix for those things that we *know* are not to |
45 |
be merged upstream anytime soon and which are /Gentoo/ specific (like some |
46 |
of our booleans). |
47 |
|
48 |
Any objections here? |
49 |
|
50 |
Wkr, |
51 |
Sven Vermeulen |