But having the ability to trace entries is useful. With something like
a public key signature, well-respected contributors could unforgably
sign their entries. Clients could then give preference to those entries
whose contributors they have been instructed to trust over ``competing''
entries. The assignment of trust has to be up to the individual user,
but the process can be almost transparent: whenever you download an
entry, you have the option of saying whether you think the quality of
this entry is especially good or bad. Gradually, your client builds
up a list of people you think write good entries. Then if a new query
returns multiple entries, your client may be better able to pick the
one you will think best.
To automate the trust mechanism further, when you see a good entry,
you could ``second'' it, adding your signature to the original author's
(or vote against it). People who trust you will then be able to benefit
from your judgement of others. I imagine that a few well-known groups
would evolve to act as database police---but rememeber that they can't
throw out entries; merely say that they think the entries are no good.
If no-one listens to them, they are ineffective. Unfortunately, this does
not protect the servers from a malicious flooding attack, but it does
go some way to protecting the clients.
To take it even further (and imitate another good idea from PGP) users
could submit their own trust lists to the database. Then, when you find
someone who you think is reliable and (transitively) you trust to make
reliability judgements, you can merge their trust list with your own.
A web of trust is quickly established. But this issue is rather beyond
the realm of current discussion and more like the subject for a PhD:-)
I presume the existing cddb solves this by having what amounts to a
moderator? Is this another single point of failure? If so, we definitely
need to come up with a workable alternative.
Robin.
-- R.M.O'Leary <robin nospam at acm.org> +44 7010 7070 44, PO Box 20, Swansea SA2 8YB, UK