Pick and the Second Law of Thermodynamics

[Wols Lists]

I doubt any of you are particularly interested in reading LWN, where
I've been making a few waves recently. But as part of my conversations
there, I had rather a brainwave ...

List the following structures in order of increasing entropy...

Matrix
List
Bag
Set

Now given that the Universe is expanding to a state of random disorder,
and this is characterised by description of "increasing entropy", it is
clear that more entropy is a bad thing. Guess which database has the
least entropy ...

So now when we have an argument about which database is best, we can
point at Physics and say "Physics proves it is us" :-)

The classic Physics explanation is Maxwell's demon, who has to Work in a
cooling unit separating hot and cold atoms. That creates information and
reduces entropy.

Cheers,
Wol

[Jim Idle]

The lowest entropy is achieved by the maximum number of relationships defined between datum. If all relationships are define between datum, then the information increases and entropy decreases -theoretically to nothing, but at any rate to the lowest for a model. 

While the result of a read of a hashed set of bytes is not quite noise, there are no relationships defined in any formal manner, and therefore entropy is higher than a more formally defined database. If all datum relationships were completely defined, then the data would effectively be redundant as all things would be known. When nothing is known, it is noise. You can stick anything you want in an item, the knowledge has to be built in to the program. 

Your argument does not hold water.

--
You received this message because you are subscribed to
the "Pick and MultiValue Databases" group.
To post, email to: mvd...@googlegroups.com
To unsubscribe, email to: mvdbms+un...@googlegroups.com
For more options, visit http://groups.google.com/group/mvdbms
---
You received this message because you are subscribed to the Google Groups "Pick and MultiValue Databases" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mvdbms+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mvdbms/89da4bd4-28b9-1c90-8410-1c47574c3f1d%40youngman.org.uk.

[Bob Markowitz]

I fully concur with Mr. Idle!

[Wol]

On 22/03/2022 05:50, Jim Idle wrote:
> The lowest entropy is achieved by the maximum number of relationships
> defined between datum. If all relationships are define between datum,
> then the information increases and entropy decreases -theoretically to
> nothing, but at any rate to the lowest for a model.
>
> While the result of a read of a hashed set of bytes is not quite noise,
> there are no relationships defined in any formal manner, and therefore
> entropy is higher than a more formally defined database. If all datum
> relationships were completely defined, then the data would effectively
> be redundant as all things would be known. When nothing is known, it is
> noise. You can stick anything you want in an item, the knowledge has to
> be built in to the program.
>
> Your argument does not hold water.

And I think you've just indulged in my favourite hobby - failing to
engage brain before shooting from the hip.

Let's compare like with like - the Pick physical model is indeed
hash->blob. Let's compare that with the Relational physical model ...
hello is anyone there ? ... crickets. Silence. There IS NO Relational
physical model to compare it with. This is why Relational is Pure Maths,
has nothing whatsoever to do with Science, and why we can make
scientific predictions about Pick in general, rather than being forced
to run experiments on every single version of every single Relational
database.

The Pick LOGICAL model is hashkey->matrix. The Relational logical model
is the set. Here we CAN compare, and the matrix contains ORDINAL
relationships galore.

Cue Maxwell's Demon working overtime putting back all the ordinality
that sets can't store.

I think you're trying to hold water in a sieve :-)

Cheers,
Wol

>
> On Mar 21, 2022 at 12:42:37 AM, Wols Lists <antl...@youngman.org.uk

[Jim Idle]

And I was pulled up for talking about Bob and his expense account. It’s pseudo-scientific nonsense. For a start an item is not a matrix - it is a sparse tree with a limited path set and no formal definition. But anyway, I have better things to do

--
You received this message because you are subscribed to
the "Pick and MultiValue Databases" group.
To post, email to: mvd...@googlegroups.com
To unsubscribe, email to: mvdbms+un...@googlegroups.com
For more options, visit http://groups.google.com/group/mvdbms
---
You received this message because you are subscribed to the Google Groups "Pick and MultiValue Databases" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mvdbms+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/mvdbms/055ae1dd-b766-f83e-1a1a-0a41582d4e00%40youngman.org.uk.

[bdeck...@gmail.com]

The universe is in chaos.  Recall Zaphod Beeblebrox and vote Idle for President!

To view this discussion on the web visit https://groups.google.com/d/msgid/mvdbms/CAGPPfg_bRA2EcpEBV1ONw-23goHT2OH-mK0prhXTd-Wt1zfGgA%40mail.gmail.com.

[Peter McMurray]

I agree Bruce The Greeks had it correct  "In the beginning all was chaos".

Then came Don and Dick.

It has long frustrated me that people do not understand the multidimensional nature of the Pick Database.

An Item can be a multi-dimensional matrix. That is the purpose of the Controlling attribute to create matrices within an Item. A feature I have used since 1977. For example I have a file debtCon and I create Items for each debtor with data such as name and credit control with  a controlling attribute of Change Date that controls Credit rating and maximum Discount Rate.

A second set is of Control Period End controlling totals for unallocated cash, opening balance, 30, 60 and 90 days or current period, period 1, period 2, period 4 depending on the period.

Also I have a range of other information  Addresses are in a second file. Nowadays I would separate the fast changing balance data into a second file for audit reasons.

Each Debtor item is a dimension of the Matrix that is called a file. The dimensions could all be stored in one item but ease of access through a sensible key that in my case is Company* Check digited debtor number.

Debtcon is in turn a dimension of the matrix called ACCOUNT that is in turn a dimension of the Array called System master.

In fact every item in a system could be stored in the system master without any further subdivision as the type of item rather like an Amazon warehouse is identified by the first attribute.

The system master account acts like a central post office the boxes are of a known size if filled they overflow to a rear storage just like a Pick Matrix.

A controlling attribute could of course be a translate to another level of information and always explodes the subject attributes. If you wish to show it alone then create a Synonym attribute that is why we have A and S types.

I believe that a common misunderstanding comes from the fact that Excel is a Row major array and Pick Files and Items are Column major array.

All data within an array is defined in the dictionary. That is why all values in an attribute must be of the same type. That is why the sub value mark was an aberration never used by Dick and why Advanced Pick trails will come to a halt if sub values are used.

Read GIRLS and Fortran arrays all is explained therein.