On 30 December 2015 at 07:07, Colin Putney <colin at wiresong.com> wrote:

>
>
> On Tue, Dec 29, 2015 at 3:27 PM, Huw Lloyd <huw.softdesigns at gmail.com>
> wrote:
>
>
>> To the degree that your conception is object-oriented, in the deeper
>> sense, it is not a problem.  Your object(s) know how to persist themselves
>> and may be considered to be extensive over that persistence -- it is merely
>> that their morphology has changed.
>>
>
> Sure, but that's just hiding the issue behind an abstraction. I'm
> interested in the mechanics of the change in morphology. If the objects can
> persist themselves, great, but what technique do they use?
>

I thought that handling a polymorphic structure was half of your problem,
Colin?  This is one of those cases where this mechanism can be delegated to
each successive object.

>
> Are you trying to identify names in a known problem space (e.g. a space
>> defined by the limits of the techniques used) or to identify the problem
>> space that you're in?  For the latter, if you're going to generalise then
>> perhaps you should also wave the specialisation of 'perfect' reproduction.
>> So, even more generally, what you seem to describing is a special case of
>> morphogenesis in which a 'copy' is made, e.g. lazy evaluation and cached
>> computation can be thought of as slices in that process.
>>
>
>  I think I'm trying to identify the problem space. "Morphogenesis" looks
> like an interesting term. Lazy evaluation and cached computation seem like
> opposites sides of the same coin. Do you mean that they are another form of
> "copying" that might be considered? I suppose that opens up new methods for
> performing the copy - eg, if the data-structure we're copying is a cached
> computation, instead of transmitting a description of the data structure,
> we might transmit the computation that produced it
>

Yes, or the 'transmit' might be pure calculation.  Really, the storing and
sending of values is a special case (short-cut) of a process.

See SICP on streams:
https://www.youtube.com/watch?v=_0i4NJYP9gM


> . Of course, then we still have the problem of how to accomplish *that*.
>

Quite so, but that is something that can be localised too, if it was
desirable.  There's no mechanistic reason why it has to be the same across
objects.

I vaguely recall that William Kent's "Data and Reality" has some of these
ideas in the form of entity - object relations which might help you too.

Best,
Huw



> Colin
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20151230/abc49f63/attachment-0001.htm