The Arrow System Paper

RE01 Rice Brian T. EM2
Mon, 3 May 1999 21:50:17 -0700

i'd just like to add some notes about ontologies:

it's rather vague to most people (those with which i've been able to discuss
this) whether ontologies are "really" arrows (as translations) or nodes (as
target models).  the former attitude is more semantically valid, since i can
define theories about a subject in a particular language that have nothing
to do with each other, or often can contradict each other.  however, as in
the example of the HLLs in the paper, i intend to represent the HLL notion
with a node, as well as the hardware state-machine notion with a node.  this
intuition led me to think of ontologies in another way: our minds
intuitively think of the target of a translation as the identifying one.
this leads to confusion when there are multiple methods of interpretation.

perhaps i am actually thinking of the process of translating the user's
ideas about, say, "how the GUI should work" into working code as separate
from the code itself (which it is).  but then we consider entirely separate
in existing systems all of the possible processes of interpreting that
language into machine state-space-time language.  so the arrow construct
obviously allows us to generalize this dichotomy into a sort of democratic
population of translation concepts.

maybe those nodes (or collections thereof) are contexts, and the information
'border' of a context can be drawn around those nodes.  then context
operations could be performed in terms of operations on that structure of
nodes and its connections with the outside world.