Hi again, Jonathan Nieder wrote:
> In its implementation of CRAM-MD5 authentication, svnserve uses > a challenge of > > < nonce . time @ hostname > > > After the fix to bug#285708, the nonce is not actually random, but is > just the current time a few moments before the later "time" component > is generated. > > After intercepting a client's response to this challenge, I can reuse > that client's credentials, provided I am fast or lucky enough to time > my connection so as to get the same challenge (nonce and time). > > The "nonce" and "time" are usually not equal and have fairly high > (1-tick --- usually 10-microsecond) resolution, so in practice this > might be hard to do. Even so, I would be happier with the following > patch applied. Have you considered this request? If the proposed fix isn't suitable, would another one be? Another detail I missed before: nowadays apr_generate_random_bytes() reads from urandom, not /dev/random, so this would not cause bug#285708 to come back. If you think the change is a bad idea, I'd be happy to tag it wontfix so it can be off my radar. Thanks, Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
