1 |
On Sun, Oct 29, 2017 at 07:47:41PM +0100, Michał Górny wrote: |
2 |
... |
3 |
> > If users need other values, it's a package-manager config knob. |
4 |
> |
5 |
> We don't want pre-EAPI times where things will fail out of the box |
6 |
> unless the user choose the one tool that got the whole list right |
7 |
> and/or configure it to account for default list. |
8 |
> |
9 |
> I don't mind package manager providing the ability to ignore additional |
10 |
> entries but the spec should work out of the box too. |
11 |
Ok, can we have a minor additions to the text then: |
12 |
- The package manager may support additional user-specified IGNORE |
13 |
entries, for usage where a user's processes need to inject additional |
14 |
files that would not be ignored by existing rules (e.g. user commits |
15 |
the rsync tree to CVS with -kb). |
16 |
|
17 |
Notes: |
18 |
- distfiles/packages/local will be in IGNORE as distributed. |
19 |
- package-manager might add lost+found if they have a filesystem just |
20 |
for the tree. |
21 |
|
22 |
> > Yes, put 'Verifying TIMESTAMP' into a new section as you added below, |
23 |
> > including the out-of-date part there; don't detail how to verify it in |
24 |
> > this section. |
25 |
> I will try to do this today. |
26 |
Looks good. |
27 |
|
28 |
> |
29 |
> > > > GLEP61, for the transition period, required compressed & uncompressed Manifests |
30 |
> > > > in the same directory to have identical content. Include mention of that here. |
31 |
> > > |
32 |
> > > Can do. But I'll do it in 'Backwards compatibility' section: |
33 |
> > > > - if the Manifest files inside the package directory are compressed, |
34 |
> > > > a uncompressed file of identical content must coexist. |
35 |
> > > > Saying that either can be used is a potential issue. |
36 |
> > > |
37 |
> > > Why? It also says that they must be identical, so it's of no difference |
38 |
> > > to the implementation which one is used. |
39 |
> > |
40 |
> > It's safe if the identical requirement is there, and potentially unsafe otherwise. |
41 |
> That's why they're both put in a *single sentence*? |
42 |
'co-exist' in this context makes it the English parse weirdly to me, |
43 |
that's why I was worried at first. |
44 |
|
45 |
Maybe a rewrite: |
46 |
An uncompressed Manifest file inside a package directory MUST exist |
47 |
during the transition period. A compressed Manifest of identical content |
48 |
MAY be present. |
49 |
|
50 |
> > > But it makes no sense when top-level Manifest is signed. This points out |
51 |
> > > that for tools not supporting full-tree verification smaller signatures |
52 |
> > > need to be used (skipping the fact that Portage did not ever implement |
53 |
> > > it). |
54 |
> > The Manifests might not be signed by the same entity. |
55 |
> > /metadata/glsa/Manifest might be signed by the security team, |
56 |
> > /sec-policy/Manifest might be signed by SELinux team, |
57 |
> > /Manifest should STILL be signed by Infra/tree-generation-process. |
58 |
> I honestly doubt this will ever happen, and even if it does, it isn't |
59 |
> really relevant to the spec at hand. My point was: if someone signs |
60 |
> the whole repository, he normally will consider it pointless to sign |
61 |
> individual package Manifests. This explains why he might consider it. |
62 |
My argument is that it make sense to permit multiple levels of signature |
63 |
even when the top-level is signed: glsa-check could get ahead of the |
64 |
Portage curve by verifying /metadata/glsa/Manifest using Gemato :-). |
65 |
It doesn't need to verify the whole tree, just that directory. |
66 |
|
67 |
The package manager should decide about the GPG-verification of the |
68 |
nested Manifests however, as they convey trust from different sources. |
69 |
|
70 |
-- |
71 |
Robin Hugh Johnson |
72 |
Gentoo Linux: Dev, Infra Lead, Foundation Asst. Treasurer |
73 |
E-Mail : robbat2@g.o |
74 |
GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85 |
75 |
GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136 |