Howard Chu <hyc@highlandsun.com> wrote on 11/04/2005 03:09:56 PM:
A couple of concerns were raised about using fractional seconds in the
timestamp portion of the CSN. Mainly, that they do not always increase
monotonically; some systems will decrement while adjusting the clock
(e.g. using NTP?). The draft-ietf-ldup-model document specifically
excluded fractional seconds; I think we need to continue with that
restriction here.
Is eliminating fractional seconds from the CSN the way to address the
problem of decreasing CSNs? Would it be more appropriate to require that
any CSN generator generate strictly increasing sequences of CSNs with
respect to the time and timeCount components? I believe there is a similar
requirement for the DCE UUID time-based variants.