Re: cdi server (mirroring ideas)

Darin Adler (darin nospam at spies.com)
Wed, 31 Mar 1999 10:40:37 -0800

>Then the question is what to do. Which server has the
>correct info? We can't take the "majority rules"
>standpoint, because the incorrect info may have propigated
>faster. So we establish a list of trusted servers. This
>parts from the "all servers being equal" standpoint, but I
>don't know how else to do it. If we take the trusted server
>route, then the CDid,chksum pair request should be to the
>trusted server base, and the chksums that differ should be
>re-requested from the trusted server and replace the
>incorrect info. And, once a CD has been varified with the
>trusted server, it doesn't need to be verified again. Maybe
>add a verified field to the record, so when it's
>distributed, the server that gets it, knows that it doesn't
>have to be verified.

One idea is to have each set of data for a CD contain a list of checksums
of "wrong" data that it supercedes. That way, a server can confirm that
someone has looked at the both sets of data and determined that one is
better than the other and distinguish this from two sets of data that
have never been compared. The data that claims to be better will then
"win" over the poorer data.

There's still a problem if someone maliciously puts bad data into a
server trying to supercede good data, and there still might need to be
some kind of "who's on top" or "trusted server" approach when two sets of
data both claim to supercede the other. But with this approach, if two
people submit different data for the same CD, only one person has to look
at the two and decide which wins. The new data will then supercede the
old on other servers without human intervention.

Thoughts on this approach?

-- Darin