[db-wg] Update of mntner object with mixed authentication

Hi Denis

On Wed, 18 Jul 2012 15:11:01 +0200, Denis Walker <denis at ripe.net> said:

> The current arrangement of hiding MD5 password hashes is based on a 
> series of community discussions and two iterations of the 
> implementation. Although the consensus is that hiding the hashes is 
> beneficial from a security point of view, unfortunately this does result 
> in some corner cases that are not easy to resolve. This is an extreme 
> example of such a corner case with so many people sharing the use of one 
> MNTNER.

> Currently there is no simple way for a user with only PGP credentials to 
> modify a MNTNER object like this one. Only one of the users with a 
> password can query the full object. Wilfried has suggested one work 
> around. Bear in mind that these corner cases only occur when there is a 
> mixture of credential options. If all users used either password or PGP 
> there is no issue. So another work around in this case could be for the 
> PGP users to included a strong password as well. As there are already so 
> many passwords in this object, perhaps this would not affect the overall 
> security level.

Yes, that's the path I've taken.

> The RIPE NCC is currently re-developing the whole of the RIPE Database 
> update software. As part of this process the RIPE NCC would like to put 
> a proposal to the community for additional authentication options 
> including an extension to the RIPE NCC Single Sign On service (SSO) to 
> cover authentication of updates to the RIPE Database. This could provide 
> a long term solution to the MNTNER problem.

> We are still in the early stages of this re-development, which we expect 
> to last for a few months. So we don't yet have the full details of 
> additional authentication options. But when we do we will submit it to 
> the community for consideration. The RIPE NCC is also always open to 
> suggestions from the community for solutions to known problems.

For the case at hand, it would be enough to have a method to
authenticate *queries* for mntner objects with any of the valid
methods for updates (not just passwords).

Regards,
Alex