Excerpts from Michael Stapelberg's message of Fri Mar 19 15:46:29 -0400 2010: > Excerpts from Eric Sherman's message of 2010-03-19 20:34:32 +0100: > > Would inserting an SHA1 into the id solve these problems, or does it just > > introduce more problems? > The latter. The ID needs to be linear, not only unique.
Oh I see. Is it fair to say that if a set of messages are can only be disambiguated by adding nanosecond precision to their timestamps, the ordering across such a set may as well be arbitrary? In looking at make_id, it seems like we should be able to simply append more numeric noise at the end of the id. The file size is already appended to the mtime, potentially making the ordering arbitrary for equivalent mtimes. Would appending x-digits of numeric noise, probably derived from SHA1, be suitable? _______________________________________________ Sup-devel mailing list Sup-devel@rubyforge.org http://rubyforge.org/mailman/listinfo/sup-devel
