Framing the issue as Bill Nitzberg did:
> 1. Is this a technically good idea?
> 2. How big a change is it?
Here is my attempt to summarize(*) where we stand, and what might
be acceptable. (Details may be found in the original mail notes.)
Regarding question 1:
There appears to be a general agreement that the change would simplify
implementations, although some doubt was expressed as to whether that
simplification was necessary.
The change also helps to simplify the user interface, although at a
cost of abandoning the "one filetype per file" (or, "one filetype per
large complex data layout") idiom for at least some applications.
Several people are concerned about the loss of expressiveness which
this entails.
Regarding question 2:
Bill N. enumerated what appears to be the complete set of
modifications to text required to implement the change. There was a
concern stated about possible subtle problems.
Next step:
There seems to be rough, but not unanimous, agreement that the "whole
filetype access" change would be worthwhile. I suggest that we move
forward cautiously, planning to make this change while giving the
possible pitfalls additional close examination - and plan to back off
if the problems start to look worse. Any thoughts?
-Ian
(*) I referred to notes from John May, Rajeev Thakur, Jean-Pierre Prost,
Sam Fineberg, Bill Nitzberg, and Dror Feitelsen