Gentoo Archives: gentoo-dev

From: Miroslav Rovis <miro.rovis@××××××××××××××.hr>
To: gentoo-dev@l.g.o
Subject: [gentoo-dev] Is it safe to switch from webrsync to the git repo now?
Date: Mon, 19 Dec 2016 23:47:40
Message-Id: 20161219234756.GA4008@g0n.xdwgrp
1 Hi!
2
3 When installing from local overlay
4 (https://wiki.gentoo.org/wiki/Overlay/Local_overlay) which I built very
5 simply from https://github.com/deuiore/palemoon-overlay (I know I could
6 have used layman and gone the regular way, but the reasons follow
7 below), and then installed Pale Moon today, but without any (obvious)
8 means to verify the git repo pulled, let alone the packs in the git
9 object dir that downloaded in /usr/portage/distfiles/,
10 I realized there seems not to have been developed a secure
11 method for the end user to update the local installation.
12
13 ( Pls. note that the particular case with the Pale Moon overlay bears no
14 importance in my query, or only as much as s single instance in
15 comparison to all instances of some method applied.
16
17 This is a question about verification of anything portage *via git* with
18 respect to simple and reliable, never failing, but obsolete method of
19 verification of portage *via webrsync*. )
20
21 I actually deliberately and kindly borrowed the title to my email from
22 this topic:
23
24 Is it safe to switch from webrsync to the git repo now?
25 https://forums.gentoo.org/viewtopic-t-1038300.html
26
27 and I can't stop wondering that nothing seems to be moving towards that
28 direction.
29
30 That topic on Gentoo Forums was started by Ant P., and seconded by, in
31 effect only one other member of the community. Looking up the Portage &
32 Programming subforum it was posted in, it has been viewed only,
33 ( currently at this address the numbers can be read:
34 https://forums.gentoo.org/viewforum-f-8-topicdays-0-start-825.html )
35 [has been viewed] only:
36
37 3159 times by the time of my writing of this (4 contributors only, Feb
38 to Jul this year).
39
40 And it's a major functionality loss, if I'm correct in my assuming that
41 nothing has been moving in the direction of finding some way to provide
42 that functionality. I'll be very glad if it turns out my assuming is
43 wrong.
44
45 I have been using webrsync-gpg exclusively for years. I also use my own
46 local Gentoo mirror, and install in Air-Gapped, and clone the master
47 Air-Gapped system onto (at least one) another same-hardware system and
48 thn I use the clone for online.
49
50 I'm construing some of the citations from that topic, into the text
51 below as if they were emails that I reply to, which they of course are
52 not.
53
54 I'm posting here these thoughts because my itch is just no different than
55 Ant P.'s and tholin's below.
56
57 Ant P. on Tue Feb 02, 2016 1:42 pm wrote:
58 > I've been using emerge-webrsync ever since it came to light the rsync
59 > repo had no security whatsoever, this was before Gentoo officially
60 > switched to git for the main tree.
61 > ...
62 >
63 > But I'm unable to find one important piece of information in the docs:
64 > the whole point of emerge-webrsync is that it checks gpg signatures
65 > automatically for me via a FEATURES flag so I don't have to go jumping
66 > through hoops to do it manually. What's the equivalent configuration
67 > option to validate commit signatures in gentoo.git, or is it already
68 > sane by default?
69
70 tholin on Mon Jul 18, 2016 10:11 am wrote:
71 > As I see it webrsync-gpg protects agains mitm attacks from the user to
72 > the mirrors and compromised mirrors. Can git do the same?
73
74 Is it really as bad as tholin in that topic states:
75
76 tholin on Mon Jul 18, 2016 10:11 am wrote:
77 > I grepped portages source to find out how it used git and I can't find
78 > anything to indicate it verifies signatures. If git is going to verify
79 > the commit signatures it also needs all the developer keys. Those keys
80 > are not part of app-crypt/gentoo-keys and I can't find any other
81 > convenient way of obtaining them. There are about 200 active
82 > developers so you'll have to hunt for their keys like pokemons.
83
84 Is it really that bad? Irreparably bad, because there is no true
85 protection against compromised sources or/and mitm attacks?
86
87 Is is really true that:
88
89 tholin on Mon Jul 18, 2016 10:11 am wrote:
90 > This only leaves the suboptimal webrsync-gpg method.
91 and there is no way to provide to the end user an equivalent method of
92 verification with git?
93
94 Sincere regards!
95 --
96 Miroslav Rovis
97 Zagreb, Croatia
98 http://www.CroatiaFidelis.hr

Attachments

File name MIME type
signature.asc application/pgp-signature