Re: Temporal database - no end date

V.J. Kumar wrote:
> Could you show with an example how the loss occurs ? You may be right, but
> let's see.

   Suppose I have 5 things. 4 of them lasted for 1 'day' before they break.The other broke on the same 'day' it went out.

   What is the mean time to failure?

     Sum of 'time quanta' = 4.
     Number of things      = 5.

    Pick any intuitive unit of 'time quanta', and you can construct an example where this problem arises.

> You've answered that yourself below: by using home-made rational numbers
> or the prepackaged floating point numbers.

   Neither of which is the approach the Date, Darwen and Lorentzos book describes.

   They have a model of time where (from pp 62) "We need to distinguish carefully between time quanta as such, which are the smallest time units the system is capable of representing, and the time units that are relevant for some particular purpose, which might be days or months or milliseconds (etc., etc.). We call these latter units time points (or just points for short) in order to stress the fact that for the purposes at hand they too are considered indivisible. Now, we might say, informally, that a time point is a "section on the timeline" - in other words, the set of time quanta - that stretches from one "boundary" quanta to the next. .... Formally, however, time points are indeed points - they are indivisble, and the concept of duration strictly does not apply."

    The argument made in response to this is to point out that regardless of the logical size of the quanta selected, there will always be applications which reason about 'fractions of a quanta' for some purpose. And this 'quantum model of time' cannot represent that kind of reasoning.

     It is telling that the DDL book limits itself to figuring out the implications of their temporal model in the WHERE clause.

[ snip ]

> But you maybe right about chronons as inadequate approximations although
> their inadequatness has got nada to do with Zeno, the continuum and other
> cool sounding nonsense.

  Then let us please dispense with the idea that the DDC book is a serious attempt to model the logical properties of time and temporal reasoning. Rather, it is an exercise in working out how a particular temporal model can be developed within the DD 3RM framework.

   And their book is worth reading for other reasons. I think they got the transaction time v application time reasoning exactly write. I think their 6th normal form discussion is useful. I think they say a bunch of interesting things about data models and time. OTOH, I think they (and Snodgrass) overlooked a big contraint question by insisting on the relational 'key' definitions built around notions of scalar quality. (Short version: If my relation is about assigning trains to tracks, what I want to capture is the idea that two trains can't share a track for overlapping intervals, and that can't be captured using keys because it involves an inequality). Received on Fri Jan 19 2007 - 23:04:15 CET