Re: Every time ..., was Re: General formula

[Ed Gerck <egerck@xxxxxxxxxx>]

Tony Bartoletti wrote:

> Ed,
>
> I still think the central problem begged by the formula is
> "What IS an attribute"?

Tony:

Yes, and in this context this is for PKIX to define. But, in the
equation, this is not relevant -- please see the comments I made
on "sub-attributes" in the original posting.  However, it may be
useful to group or ungroup attributes in order to simplify
calculations.

> To take Alan Lloyd's example, I certify a color table.  I can
> (it would seem, arbitrarily) consider the entire array of values
> as the table's "single attribute", or consider each color entry
> (say RGB triplet) to be an attribute, or even take each R, G,
> and B value separately as an attribute.
>
> If the lifetime/lifespan of the table depends upon the lifetime
> of each individual component, how do I know at what level of
> granularity to assign attribute-hood, and thus apply the formula
> correctly?

If you consider the whole certificate as one attribute -- fine. Then,
there is few to calculate but there is all to model.

If you consider each R, G, B component as different attributes -- fine.
Then, there is all to calculate but few to model.

Engineers usually work in-between these cases. The equation applies
to either case and also in-between.

> If you certify my name as "Tony", is that one attribute,
> or is it four (each letter a separate attribute.)
>
> As a first guess, I would say that if something can change
> independently of another thing (and yes, are treated semantically
> as separable values), then they can be separate attributes.

Yes.

> Otherwise, _in_the_context_ where two things are treated as
> completely dependent (redundant) by definition, then they cannot
> be treated as "two", as they are really "one".

Yes -- but most things are neither. Most things are partially dependent.
However, even if they are partially dependent, the equation is still the
same because the equation deals with the *effective* validity for each
attribute. I tried to explain this in the inlined reference [1] of my original
posting -- it would be useful if you could please check it and verify if
it is clear (at least, in hindsight).

Cheers,

Ed Gerck