[Cuis-dev] About complexity in the face of the user

[ken.dickey at whidbey.com]

On 2024-03-19 14:32, Hilaire Fernandes via Cuis-dev wrote:
Le 17/03/2024 à 22:22, ken.dickey at whidbey.com a écrit :
My thought was to look at blocks languages 
(Scratch/BYOB/GPBlocks/Blocky/..) and make a small blocks language based 
on DrGeo construction semantics (objects, operations, 
composition/extension).

Yes, may be. What I have in mind is how you will program a given
model, not design it, but just make use of it by programming it. This
will result in using both general programming instruction and
dedicated ones to the model to interact with. The most obvious -- easy
for us -- is to use the Smalltalk language, after all this what it was
designed for. Of course you can start to parametrize the model, but
this is not really programming.

I guess my "half baked" idea is "training wheels".  Like training wheels 
for a bicycle, one could have training wheels for domain specific 
construction language(s) leading to Smalltalk.  Note: Pencil graphic 
(attached).  Lousy syntax but interesting idea.  For visual learners 
like myself (linguistic/parser/algorithmic ideas take me longer to 
puzzle out than visual metaphors) a Blocks Language is an easy entry to 
usage of loops and events and it is easy to transliterate a Blocks 
program/script into Smalltalk.

The trick in all of this is to make a natural/easy transition between 
the visual "construction kit" aspect and the larger programming universe 
where one can do much more interesting things but at greater risk of 
breakage.

I found an interesting thumbnail discussion in "Blocks and Beyond" ( 
https://arxiv.org/pdf/1705.09413.pdf ).

DrGeo is my test bed to explore these ideas:  the DrGeo Smalltalk
sketch editor (you see picture in my previous email), with its
dedicated editor and resulting view, then my exploring of debugging
with the mini-debugger.

Yes. I need to spend more time with DrGeo.  At the moment I just adopted 
a "rescue dog" and he is chewing on my foot, so I am a bit distracted. 
Ouch! :^)

What I am really interested is how you will go beyond DrGeo, and have
programming capabilities for any kind of model written in Smalltalk,
then the user can make these models to interact with each other.

Yes.

One observation.  It used to be the case that one could walk through the 
local village and see a full range of skill levels.  Shoes, furniture, 
clothing being made, blacksmithing/fabrication of various sorts.  People 
doing apprenticeships to learn trades. Today we basically see only "raw 
beginners" and "Olympic quality athletes" with very little in between.

If you get someone untrained with my kind of psychology, training wheels 
are useful to lead me from exploring/tinkering to 
planning/strategy/algorithms/pattern-abstraction.

So I am thinking of introduction, refinement, iteration to "bridge" me 
toward goals rather than working on the ideal endpoint.

Yeah, this really sound as reinventing Etoys, but this is not what I
am interested by. My metaphor is more like a notebook where you
handwrite notes and insert model a user will have scripted (with
Smalltalk or tile), alternatively the user will retrieve programmed
model  from a library of scripts. Think about a set of elementary
sketches representing different king of interactive parallelogram. You
could why would you need script for, it could be just DrGeo sketch.
Well what is nice about script is your can learn from, and it may be
easier to modify and to extend.

Right.  [I think we are probably "agreeing loudly" here ;^].

I am just trying to get there with my odd little brain.  You have a 
larger universe of teaching/learning puzzles to manifest.

Good on ya,
-KenD


-------------- next part --------------
A non-text attachment was scrubbed...
Name: Pencil.png
Type: image/png
Size: 212019 bytes
Desc: not available
URL: <http://lists.cuis.st/mailman/archives/cuis-dev/attachments/20240320/fc6b72ee/attachment-0001.png>