[sc34wg3] TMRM v6.0 comments

[Patrick Durusau]

Nikita,

Yes! Please continue with your comments!

Further replies below.

Nikita Ogievetsky wrote:

> Dear Steve, Patrick, Robert, and all,
>
> I really like the direction in which TMRM is moving, very impressive!
>
> I can not say that I understood everything, there are some 
> questions/comments and some potential reservations (which are probably 
> due to the open questions :-)).
>
> I’ll start with questions/comments:
>
> 1)
>
> Page 4:
>
> “Proxies and their identifiers are interchangeable”
>
> What does it mean and how can that be?
>
> It contradicts with:
>
> “Proxy is a set of finite set of properties” (p4)
>
> “Subject proxies could have the same key/value pairs, but different 
> identifiers” (p9)
>
(The following is not entirely free from doubt or disagreement.)

The problem arises because proxies in the most abstract sense are 
representatives of subjects and have no "identifiers."

But, in order to do anything with proxies they have to be represented in 
an information system, which gives an instance of a proxy an identity 
vis-as-vis the system in which it is appearing. That identity does not 
have anything to do with the subject that it represents.

In other words, in an information system, a proxy has two separate and 
distinct identities, 1. as a representative of a subject, and 2. as an 
addressable instance in the information system.

I think (without having asked) that everyone would agree that a system 
identifier does not have any relationship to the subject that is being 
represented by a proxy. So, it appears to me that system identifiers 
should not be considered as a property of a subject proxy.

But, it is equally true that being able to reference subject proxies by 
a system identifier is a very useful.

The problem comes in, as you point out, what does it mean to say proxies 
are "identical?" If they have different system identifiers, then it 
seems the system regards them as distinct. (second sense of identity, as 
an addressable instance in the system) But if that system identifier is 
not a property of the proxy (and I don't think it should be) then it 
cannot serve to distinguish the proxies as per the definition of proxy 
in the TMRM. (first sense of "identity," i.e., the subject it represents)

I would note that all this difficulty is caused by introducing the 
special case of all properties are identical, which is included in the 
general case of equivalence. (Newcomb pointed this out to me yesterday.)

If we simply say that equivalence is declared by disclosures and leave 
the management of such cases up to disclosures, I think we would avoid 
some of the conceptual difficulty that you note in the current draft.

That seems to answer any concerns about data redundancy (the reason for 
the "identical" properties statements) about as simply as it can be 
answered.

> 2) Why introduce notion of “unordered sequence”? (p10)
>
> Instead of using old “set” or “bag”?
>
> Tastes like fat-free butter… :-) Or may be it is my English…
>
I would have to defer to Robert for a definitive answer. I suspect that 
"unordered sequence" is present simply to support the notion of "ordered 
sequence."

Bags can be ordered or unordered so it may simply be multiplying terms. ;-)

> 3) It is not clear what is in (…) between (v1j, v2j,…,v1j) in 
> expression (9) on p11
>
I think that is a typo, should be (v1j, v2j,...,vnj) which would simply 
represent a tuple sequence.

Hope you are having a great day!

Patrick

> Well… it is late now... If these comments are useful I will continue 
> tomorrow.
>
> Kind Regards,
>
> --Nikita
>

-- 
Patrick Durusau
Patrick@Durusau.net
Chair, V1 - Text Processing: Office and Publishing Systems Interface
Co-Editor, ISO 13250, Topic Maps -- Reference Model
Member, Text Encoding Initiative Board of Directors, 2003-2005

Topic Maps: Human, not artificial, intelligence at work!