Re: Overlay mechanic improvements

[Stephen J. Turnbull]

Richard Stallman writes:

 >      > But you keep the overlay in existence -- how come?
 > 
 >     Why do you care?
 > 
 > Because I want to understand what this program is doing.

That much is obvious.  But why do you care what the program is doing?
Do you plan to become an AUCTeX developer?  Do you want to check
whether it's possible for AUCTeX to use text properties rather than
overlays?  Do you want to check if AUCTeX is highly wasteful of
overlays, leading to the efficiency issue, so that an improvement to
Emacs overlays is actually unnecessary for the use case?

The first motivation seems implausible given your time constraints and
non-use of AUCTeX, and I am quite sure the second two are wastes of
your time -- you will discover that AUCTeX's overlay implementation is
preferable to using text properties and that it doesn't leak overlays,
it just uses a lot of them in a big document.  For two reasons: I know
a little bit about the implementation, and independently come to those
conclusions, and as confirmation, I know that David knows his business.

Do you have yet another motivation for studying AUCTeX's implementation?

 >       David knows his business, isn't that good enough?
 > 
 > Good enough to tell me the program logic???

Of course not.  But knowing the program logic tells you what it is
doing, not why it does it that way.  If the program is a poor program,
you could probably find ways to optimize it, and so avoid a large
change to Emacs.  But I see every reason to believe that is not the
case.

At this point, from my point of view, we've been going around in a
tight spiral for several days.  It would be nice if we could go
straight to the point, that's all.  Or, it would be nice if you could
show that we are already on the straight path to the point.