[db-wg] Filtering auth: attributes in the whois server (was: Re: Suggestion...)

Hi,

I'm following up on the "how to get at your complete maintainer
object" case which was the subject of this thread.  I've had a
private discussion with the people manning the ripe-dbm at ripe.net
role, and this is the information which has been revealed so far:

   The current implementation of the filtering in the RIPE whois
   server is such that *all* auth: attributes are removed from all
   maintainer objects, as we have all seen.

   As a LIR in the category who only use PGP auth: lines in our
   maintainer object (we're one of the 16% of all maintainers who
   are in the category "have either not used MD5-passwords or
   have abolished doing so"), I may still get at my full
   maintainer object without using any password.  However, it has
   been confirmed that now I *do* have to use the web interface.
   I have to visit

     http://www.ripe.net/webupdates/

   and activate the "Modify object in single text area" option
   before looking up my maintainer object.  I'll then have to cut
   and paste this into whatever I use to update my maintainer
   object (in my case: my e-mail client).

   So...  It is confirmed: this is forcing us to use the web
   interface to look up our maintainer object, where we could use
   the whois protocol earlier, something I'm unhappy about.

   I suggested to the RIPE NCC that instead of following the
   current implementation strategy, they could have followed the
   following strategy:

     In the whois server, if a maintainer object *only* contains
     auth: lines using currently deemed to be secure methods
     (currently PGP or X.509), then reveal all the auth: lines to
     the whois client.  Otherwise, if the maintainer object
     contains one or more lower-security auth: attribute
     (currently MD5-based passwords), filter out *all* the auth:
     attributes.

   This would not require a behavioural change from those LIRs
   who have "done the right thing" and abolished MD5-encrypted
   passwords (or never even used them), instead of punishing us
   by forcing us to use the web interface when updating our
   maintainer objects.

   The complication that the RIPE NCC sees with this solution is
   that this would make it more difficult to explain what
   procedures need to be followed to update a maintainer object.
   My claim is that whois + e-mail if you do the right thing
   (only use PGP or X.509) and use web-updates if you do the
   wrong thing and use MD5 passwords is no more difficult to
   explain than the current situation, and as a bonus it does not
   punish LIRs who have already done the right thing and
   abolished MD5-based passwords.

The RIPE NCC also basically said "this is what the community
agreed to at the RIPE-63 meeting", and that if a change should
come, the community should agree to it.

Therefore I am bringing this matter to the RIPE database working
group for discussion.  My proposal is the alternative implementation
strategy quoted above.

Best regards,

- Håvard