Re: what are keys and surrogates?

From: Bob Badour <>
Date: Fri, 18 Jan 2008 13:33:13 -0400
Message-ID: <4790e2de$0$4042$>

Jan Hidders wrote:

> On 18 jan, 16:55, Bob Badour <> wrote:

>>Jan Hidders wrote:
>>>On 18 jan, 00:55, Bob Badour <> wrote:
>>>>Jan Hidders wrote:
>>>>>On Jan 17, 10:29 pm, Bob Badour <> wrote:
>>>>>>Jan Hidders wrote:
>>>>>>>On 17 jan, 14:40, Bob Badour <> wrote:
>>>>>>>>Marshall wrote:
>>>>>>>>>On Jan 10, 7:07 am, Bob Badour <> wrote:
>>>>>>>>>>Marshall wrote:
>>>>>>>>>>>>Is "constructor" the same as what C. Date calls a "selector"?
>>>>>>>>>>>Yes. Date calls it a selector, and the entire rest of the world
>>>>>>>>>>>calls it a constructor. :-)
>>>>>>>>>>Except "selector" has no concept of physically building anything in storage.
>>>>>>>>>Okay. Just specifying a value, or a kind of value, yes?
>>>>>>>>>That's more or less what I understand the most general
>>>>>>>>>definition of the word "constructor" to mean. The OOP
>>>>>>>>>world uses it a bit more specifically.
>>>>>>>>I suspect the word originates in the OOP world, and it strongly suggests
>>>>>>>>building something physical.
>>>>>>>>>>>I have no strong feelings about encapsulated ADTs; what
>>>>>>>>>>>Date calls ... uh. Shit. I can't remember what he calls them.
>>>>>>>>>>>I don't entirely see the reason for them. Performance I guess?
>>>>>>>>>>Types? Possible representations? Type generators? Only the first is an
>>>>>>>>>>ADT, but I am curious whether you meant one of the others.
>>>>>>>>>Possreps! That's the one!
>>>>>>>>Having multiple possible representations for the same type allows data
>>>>>>>>independence--especially physical independence.
>>>>>>>You think the number of possible representations and the number of
>>>>>>>possible ways to store something in memory or on disk are related?
>>>>>>The latter grows linearly with the former.
>>>>>Why? Seriously. Why do you think there is a relationship at all? Why
>>>>>would the number of ways a value can be represented to the user
>>>>>(something which a matter of definition and/or convention) have any
>>>>>bearing on how many ways there are to map it to 1's and 0's?
>>>>Why do you think representations are limited to representing to users?
>>>I don't. I think possreps are. But I could be wrong about their
>>I strongly suspect you are.
>>>So what is exactly a possrep according to you? Are you
>>>saying that the number of possreps of a certain value is by definition
>>>the number of ways it can be represented as ... as what?
>>I don't understand the part starting with "as". I would punctuate the
>>sentence there instead.
> I think you want to be more specific on what a representation exactly
> is. That it should be something that can be presented by a computer to
> a user on screen or a paper, for example. 

A poor example.

A string, would be my first

> guess. Other representations like smoke signs, chalk marks, a series
> of coughs, etc. are clearly not very relevant for the question in how
> many ways you can map it to the physical layer.

I disagree.

>>>>How does representing to a user differ from representing to a machine?
>>>You don't get out much, do you? :-)
>>That kind of man-machine interface never even occured to me. One wonders
>>what causes your mind to wander in that direction.

> I was only half kidding. A possrep is a notion at the logical level,
> and therefore a matter of agreement between people. It is how a
> certain set of persons agree to communicate a certain concept. For a
> mapping to a physical level there is no such agreement necessary.

Indeed. However, every mapping to the physical level starts with a possible representation. Double or triple the number of possible representations, and one doubles or triples the number of possible physical representations.

Data independence is not limited solely to physical independence, either.


> makes these notions fundamentally different. More to the point, you
> could have six ways to represent the number 5, but let the DBMS store
> it always in the same way.

But one would have to choose a possible representation. Choosing a representation that does not possibly represent the value seems pointless.

  Or you could have only way to represent it, > but have six ways of storing it.

If one has six ways of storing it, presumably one has six ways of storing each of the possible representations. Thus, the number of physical representations grows linearly with the possible representations.

  There is no direct relationship.

You appear to lack imagination.


> best you could say that if there are multiple representations, this
> helps the system to realize that it might store it in different ways,
> viz., the naive mappings of these representations, but the latter
> notion indicates a link between logical and physical notions that in
> principle should be there.

Anthropomorphizing machines regresses the conversation to nonsense. Would you care to rephrase the above into something sensible? Received on Fri Jan 18 2008 - 18:33:13 CET

Original text of this message