Re: Impossible Database Design?
Date: Sat, 20 May 2006 01:28:42 GMT
Message-ID: <ehubg.35069$mh.17481_at_tornado.ohiordc.rr.com>
-CELKO- wrote:
> I cannot shake Zeno's paradoxes which occur with a discrete model of
> time.
See, now I thought that to be a consequence of a continuum in that it may be divided infinitely. And I thought the concept of concrete mathematics bridged the gap for those of us who, because we use computers as general-purpose tools, have to work in the discrete world.
>>>I'm not sure if I am getting this the way you mean it. You mean p1, p2 etc. as time points? <<
>
> NO, P# as part numbers! He uses the [<start> : <terminal>] notation
> for anything.
Well, not anything. (More to follow...)
>>>During Birds [p1:p5] pink lawn flamingoes <<
>
> Exactly! See what I mean about how it does not make sense! Then they
> have PACK() and UNPACK(), etc.
>
>>>I'm still wondering wether my interpretation of your questions is correct. <<
>
> Yes, and it really is that crazy for a set of unordered distinct
> elements whose names can be put into a sorted order. It crossses
> logical and physical bounds and makes no sense.
That may be why you're having difficulty. DD+L specifically excluded a continuum. The domains for which their interval operators work are necessarily comprised of discrete values for which predecessor and successor operators will return a value in that domain. Using bits of your example,
red flamingo < pink flamingo < white flamingo may be the domain and
next(red flamingo) = pink flamingo = previous(white flamingo) is defined. You don't like pink? Okay. Change the domain. Of course, you have to paint all the pink flamingos in existence -- but maybe there aren't any. Your idea of eliding Tuesday has nothing to do with their proposed technique; it's patently absurd. (As would be yellow flamingos! But folks in Austin have all sorts of strangeness on their lawns)
Regarding domains: the question, "Is absorption a property of some value in the domain?" must be answered depending on what the domain is and how it can reasonably be represented. If, as someone else said in this thread, "all time intervals have a starting point but they may not always end" (Or something to that effect. Seems strange to me, but...) one may want to use something they call NULL to represent the end of time which will never occur. Then, next(NULL) = NULL and infinity has been represented. The question, "What does previous(NULL) give me?" is up to the domain modeler. If he thinks he can put a date/time stamp on the instant before a time which will never come: go for it!
DD+L were also very careful about the types of intervals for which their approach worked. And that's not all; there was more.
Overall, I believe they carefully qualified the techniques they described -- so well, in fact, that your misunderstandings indicate that you aren't very familiar with the material. Received on Sat May 20 2006 - 03:28:42 CEST