Re: Idempotence and "Replication Insensitivity" are equivalent ?
From: Marshall <marshall.spight_at_gmail.com>
Date: 22 Sep 2006 09:36:20 -0700
Message-ID: <1158942980.665658.64290_at_k70g2000cwa.googlegroups.com>
> Now supose we are talking about sets (and require A to be associative,
> commutative and idempotent), the it is clear that you can still give a
> triple (E,S,A) such that it computes for example the top 5 of the set
> according to some ordering.
Date: 22 Sep 2006 09:36:20 -0700
Message-ID: <1158942980.665658.64290_at_k70g2000cwa.googlegroups.com>
Jan Hidders wrote:
>
> Now supose we are talking about sets (and require A to be associative,
> commutative and idempotent), the it is clear that you can still give a
> triple (E,S,A) such that it computes for example the top 5 of the set
> according to some ordering.
Something that's always bugged me about this sort of construction, is that I never see authors distinguishing between a partial order and a total order. It seems to me that if the supplied ordering is a partial ordering, and you want to do something like top 5, you have to either introduce nondeterminism or change top 5 to top at least 5. Yes? Or is this just considered so obvious that no one bothers to make it explicit?
Marshall Received on Fri Sep 22 2006 - 18:36:20 CEST