Re: Ping: dawn, some mvl questions

Keith H Duggar wrote:

[snip]

> If my grandmother hands me a recipe written on a piece of
> paper and I type it into emacs to store as a text file, do I
> lose information?

Retorical, but I'll answer any, just to be specific: Yes. Historians may come up with more, but here is some: Her handwriting. Some of the layout.
The age of the paper. Dents from writings on other papers.

> What if some of the instructions were
> written heavily or traced over many times (ie bold) do I
> lose information if I use a plain text file? What if I sort
> the ingredients by quantity when she did alphabetically?

If she wrote it down alphabetically you don't lose any information by sorting the ingredients by quantity. You can sort it back alphabetically anytime.

However, if she wrote it down in a particular order, not based on the content, that order is exactly the information you lose by sorting the ingredients by quantity. You will never be able to reconstruct the source order if you don't keep it e.g. by adding the original line numbers in your plain text file.

> Of course no two physical representations (in the example
> text file versus paper) EVER retain exactly the same
> information since there is an vast amount of information
> held in any physical system (most of it unknown in fact)
> that is unique.

So let's lose the order? I don't think that is wise always.

> This is why it is so important to define logical (not
> physical!) models of data. Because we are free to define the
> logical model as we see fit and abstract away physical
> details that we deem insignificant. The word deem is key
> here. It is a choice. Our model is a choice. As long as we
> agree on the logical data model then we are fully capable of
> communicating information without loss.
>
> Perhaps my grandmother assumed I would recognize the heavily
> traced ingredients as important and the others optional.
> Perhaps I just thought she was doodling. Regardless the flaw
> is not the logical data model it is her and my failure to
> AGREE on a COMMON data model. The flaw is in
> no overt statement of the meaning of the order and
> information never verbalized
> (though "verbalized" seems a poor choice of words perhaps
> for "modeled" or "represented"?)
>
> If you want to use a logical data model that assumes some
> built-in implicit (physical?) order is meaningful nothing is
> stopping you. Though it is certainly not necessary to do so
> (the RM for example) and many have argued quite reasonably
> that such order dependence is a bad idea.

I have only seen a priori dismissal of meaning of order, coerced by the toolkit. Nails and hammers. Maybe you can reiterate the reasonable arguments.

> Next time I will have my grandmother handwrite her recipe in
> XML (blah! not! :)

Heh. The order /would/ be kept even if it isn't significant. Nails and hammers again, but now working in the opposite direction. Here we agree. Received on Mon May 22 2006 - 01:15:39 CEST