On Sun, 18 Sep 1994 Jueneman(_at_)gte(_dot_)com wrote:
Now suppose that the FCC requires that Competitive Access Providers must be
allowed to provide local exchange telephone service to residential users
within a given city on a nondiscriminatory basis. This might result in one
house being served by GTE and its neighbor being served by PacTel. Worse
yet, people might do what I do -- since I have two incoming lines, I
assigned one to MCI and the other to AT&T, hoping that a common catastrophy
wouldn't knock out both of them at the same time. Now who do you go to to
find a person's telephone number? Do you have to look in two or more
directories, either paper or electronic?
It's worse than you think. Statistics in Australia show that the percentage
of unlisted telephone numbers is on the increase. I suspect that trends
world-wide are the same.
And it gets worse: if I know you live in California, but I live in Australia,
who the heck do I call for directory assistance? Currently, I would call up
Telecom Australia and say "I'm trying to locate a number for a Californian
resident. Can you help me?" With any luck, they'll tell me "Well, we don't
know, but if you call this number, they might be able to help you".
Maybe this is what PEM's certificate directories need. Some central location
(or locations) which say "we don't know, but that site over there probably
does". This could be a function of the IPRA. It knows the names of all the
PCA's it has signed, and the PCA's knows the names of all of the CA's they
have signed. So eventually everything in the hierarchy of trust can be
traced down from the IPRA. With suitable caching and mirroring to relieve
the burden on the IPRA's server, this may even work.
For those of us that prefer non-hierarchical trust, separate servers could
be run by the IPRA or volunteers. The clients would search the hierarchy
first and if nothing turns up, try the non-hierarchical servers.
Maybe X.500 already does this (I'm pretty certain that the DNS does). I'm
no X.500 expert. But I'd probably vote for a lightweight protocol that can
be implemented quickly anyway.
The most important thing however is to have a small handful of Internet
addresses that can be configured into clients by developers so that this
searching can happen "out of the box". After initial installation, the
client starts to cache its own copy of the hierarchy starting from the
given addresses. Currently, I have no idea what address I should
configure into my clients as the "X.500 root server" or whatever.
Cheers,
Rhys.