1 |
TL;DR: Yes! |
2 |
|
3 |
* Hanno Böck schrieb am 27.03.15 um 15:33 Uhr: |
4 |
>Hi, |
5 |
> |
6 |
>Right now a number of Gentoo webpages are by default served over http. |
7 |
>There is a growing trend to push more webpages to default to https, |
8 |
>mostly pushed by google. I think this is a good thing and I think |
9 |
>Gentoo should follow. |
10 |
> |
11 |
>Right now we seem to have a mix: |
12 |
>* A number of webpages default to http and have optional https |
13 |
> (www.gentoo.org) |
14 |
>* Some with sensitive logins are already https by default (e.g. |
15 |
> bugs.gentoo.org), but they don't use hsts, which they should |
16 |
>* Some with logins are mixed http/login-via-https, which makes them |
17 |
> vulnerable to ssl-stripping-attacks (e.g. wiki.gentoo.org) |
18 |
> |
19 |
>I'd propose the following: |
20 |
>* Make all pages under .gentoo.org https by default |
21 |
>* Make sure all use modern HTTPS features, including: |
22 |
> * OCSP Stapling |
23 |
> * HSTS |
24 |
> * A secure collection of cipher suites |
25 |
|
26 |
-> bettercrypro.org |
27 |
|
28 |
> * (one may add HPKP here, but it requires careful planning and has the |
29 |
> potential to lock people out of the page if done wrong) |
30 |
>(On the long term I think it would also be good to have downloads over |
31 |
>https, but I'm aware that this is more difficult as it involves mirror |
32 |
>operators that are not under direct control of gentoo infrastructure.) |
33 |
|
34 |
+1 |
35 |
|
36 |
> |
37 |
>As I know these discussions, I'll already answer to some |
38 |
>counter-arguments that may come up: |
39 |
> |
40 |
>"It's not neccessary to do https on pages without logins" |
41 |
>These kinds of arguments show a fundamental misunderstanding of what |
42 |
>https does. It guarantees confidentiality *and* integrity. In short, it |
43 |
>protects content not only from observation, but also from manipulation, |
44 |
>which is always a good thing. A very practical example is that on some |
45 |
>networks foreign ads get injected into other peoples webpages. |
46 |
|
47 |
ack |
48 |
|
49 |
> |
50 |
>"Makes things slower / servers can't handle it" |
51 |
>The performance costs for TLS on a server are often vastly overstatet. |
52 |
>The performance hit on servers doing https is very close to zero, it |
53 |
>just doesn't matter much. |
54 |
>There are some latency problems for connections, but these can mostly |
55 |
>be wiped out by a sane configuration of the server. If http/2 is used |
56 |
>one can even improve the performance with https. |
57 |
|
58 |
And often a too slow /dev/random is the cuplrit which can be fixed |
59 |
by using haveged. |
60 |
|
61 |
> |
62 |
>"Certificates are too expensive" |
63 |
>Gentoo already has certs for all pages, so this is not an argument |
64 |
>here, but if this ever becomes an issue there are a number of CAs these |
65 |
>days that issue free certs. In summer the community based CA Let's |
66 |
>encrypt will start which will be another option. |
67 |
|
68 |
Or CAs which offer a "Cert Flatrate" for a small fee per year like |
69 |
StartSSL.com |
70 |
|
71 |
> |
72 |
>"CAs are bad and the whole system is broken" |
73 |
>Partly true, but it doesn't get any better if people stick to HTTP. |
74 |
>Many problems of the CA system can be mitigated by modern technologies |
75 |
>like Key Pinning and Certificate Transparency. |
76 |
> |
77 |
>I think defaulting the net to HTTPS is a big step for more security and |
78 |
>I think Gentoo should join the trend here. |
79 |
|
80 |
... DNSSEC with TLSA records comes to my mind |
81 |
|
82 |
|
83 |
-- |
84 |
0x35A64134 - 8AAC 5F46 83B4 DB70 8317 |
85 |
3723 296C 6CCA 35A6 4134 |