Re: Of Interest

[Niclas Olofsson <gurun@acc.umu.se>]

I reply to both your mails in this one. Quotes where neccessary.

Bit offtopic, but figuring HumanML will become a HLAL for some domain it
might be interesting anyway.

"Bullard, Claude L (Len)" wrote:
> BTW:  do you have a definition for "usability" that
> you would apply in this context? 

Yes I think I do. It would be usability in the sense of the optimal
level of abstraction, making most sense to the user, with the least
perception effort (from the user). The definition is not as precise as
one could wish for, so if I may elaborate around this a bit. What I mean
it to mean (eh), has its foundation in the basics theories about the
relation between data, information and perhaps even knowledge.
Percpetion effort, or the effort it takes to recognize and form
information from a set of data, is vital. Think math, think
simplification of functions. Think macros. The less the effort, the more
friendly to the user. Sometomes more data makes information more
perceivable (pictures), sometimes less (1+1=2 is most often easier than
"one plus one equals two").
Making sense to the user. This is a constraint telling us that we can't
just form any information from any data. It has to be recognized by the
user as information, foundation for knowledge, thus in the end make
sense.
Abstraction, or the borderline between information and knowledge.
Abstraction in the sense that information comes together to form
knowledge, to capture a surtain domain. 

An optimal (or perfect for the domain) HLAL is what I then mean is the
optimal level of abstraction over a set of domain dependend data,
expressed in way as easily perceivable as it possible can, still making
sense to the user (assuming he has knowledge about the domain in
question). Sometimes it is expressed as frameworks or concepts for a
specific domain. In XP (and some other agile methods) they use
metaphors. In the end its about "money", it's about providing highest
possible quallity, most possible features, with least amount of cost.

My definition does lack a couple of constraints though (and I'm not in a
mood to chase them down right now). The domain of Human Animation is
animation of human characters. Right now it is not HLAL over VRML
because even though it captures parts of a physical model of a human, a
H-Anim avatar doesn't really need to make sense to be H-Anim. In
programmers terms I would put it like "it doesn't have a main()", if you
know what I mean ...

Acctually the story is that when I first turned to XML it was because I
needed to produce a HLAL over VRML and didn't want to write my own
parser (that was what XML was for most of us at that time). I accutally
managed quite well, turing 4 days of high cost resource usage down to
"anybody in 5-6 hours". After that I was pretty much sold on XML, even
though most I know would have said for the wrong reasons. Shame on them
:-)

"Bullard, Claude L (Len)" wrote:
> author sees 
> 
> <partnumber>1-111</partnumber>
> <partname>Wrench</partname>
> 
> the programmer sees
> 
> <data value="1-111" type="number" />
> <data value="Wrench" type="string" />

As far I can see (I'm sure this was a very quick example) the last
example does not provide enough data to form the same information and
knowledge as the first. I know it is a number, but I can't know what it
means. It would take me a lot of effort to make me understand it. But
nevertheless, I would deduce that no matter you are a programmer or a
author, assuming they both capture the same domain, I think the first is
HLAL for the domain, and the second is not. The second could however be
HLAL over XSchemas, parts of a hashmap model or something.

I think one can reason that MMTT is not a good starting point for HLAL
production but a neccessary "evil" to be able to find the optimal level
of abstraction (since MMTT in a way always describes yet another level).
MMTT is sort of like the infinite level of abstraction. With XP user
stories or feature driven development in my back, I even think I can
proof that to be correct :-)

I also think I could be making no sense at all :-P

Cheers,
/Niclas
-- 
Niclas Olofsson - http://www.ismobile.com
Product Development, isMobile, Aurorum 2, S-977 75 Luleå, Sweden
Phone: +46(0)920-75550
Mobile: +46(0)70-3726404