Re: Additional requirements for URCs?

Daniel W. Connolly (connolly@hal.com)
Tue, 16 Aug 1994 10:40:13 -0500

Message-Id: <9408161540.AA24872@ulua.hal.com>
To: "Jon P. Knight" <cojpk@lut.ac.uk>
Subject: Re: Additional requirements for URCs?
In-Reply-To: Your message of "Tue, 16 Aug 1994 08:43:09 BST."
<Pine.3.05.9408160807.A10994-b100000@suna>
Date: Tue, 16 Aug 1994 10:40:13 -0500
From: "Daniel W. Connolly" <connolly@hal.com>

In message <Pine.3.05.9408160807.A10994-b100000@suna>, "Jon P. Knight" writes:
>On Mon, 15 Aug 1994, Daniel W. Connolly wrote:
>> If the names are really independent of the lookup servers, then you
>> can use what amount to random numbers, if you like. For example, use
>> DCE uuid's, or 128bit truly random numbers (generated ala PGP keys).
>
>If the names are random numbers and you don't use a shared lookup service
>(such as DNS), how can you be sure that the number you pick for you're new
>document/service is unique?

With DCE uuid's, they're a function of "time and space," i.e. it uses
your network address and the current time and some other things to
come up with a unique number.

With 128 bit random numbers, you settle for odds of like 1 in 10^-15
that there will ever be a collision (even with the birthday problem,
128 bit random numbers is enough to index all the atoms in the universe
with very little chance of collision.)

Dan