| 1 |
It sounds like a lot of work has been done on this. |
| 2 |
|
| 3 |
My biggest concern is that this seems to be ignoring the projects that |
| 4 |
are already signing their packages. I know that many projects already |
| 5 |
sign their packages (check out Apache, MySQL, etc) and I believe that |
| 6 |
your proposed implementation is disregarding these signatures and the |
| 7 |
existing web of trust that they represent. |
| 8 |
|
| 9 |
With your way of doing it, if I'm understanding correctly, you're |
| 10 |
ensuring that the ebuild we download is as the ebuild author intended |
| 11 |
it. But, are we sure that the tarball that is downloaded by the ebuild |
| 12 |
is as the package creator intended it? Are we checking that Apache |
| 13 |
signed the package that is being downloaded? I know an MD5 sum is |
| 14 |
checked, but where does that come from? Does the ebuild maintainer |
| 15 |
get that from the same site that the tarball comes from? What if the |
| 16 |
tarball is trojaned and then the md5 sum is replaced? |
| 17 |
|
| 18 |
Is there a way to incorporate this existing method instead of |
| 19 |
replacing it or ignoring it? The reason it's called a Web Of Trust |
| 20 |
is that existing keys can be used to verify new,unknonw, keys. By |
| 21 |
disregarding the existing practice of signing distributed packages, |
| 22 |
I'm not sure that this implementation is truly using that existing |
| 23 |
practice that works when utilized properly. |
| 24 |
|
| 25 |
Otherwise, great work! I'm looking forward to using this on a daily |
| 26 |
basis. |
| 27 |
|
| 28 |
jeremy |
| 29 |
|
| 30 |
-- |
| 31 |
Jeremy Kelley <jeremy@××××××××××.com> |
| 32 |
Information Security Analyst |
| 33 |
Jeremy's opinions are not the official opinion of IBM. |
| 34 |
|
| 35 |
-- |
| 36 |
gentoo-hardened@g.o mailing list |
Archives