[gclist] Name that hypothesis

Nick Barnes nickb@harlequin.co.uk
Tue, 03 Dec 1996 17:57:19 +0000

A set of hypotheses:

1. most objects die young.
2. old objects will not die soon.

3. most references are from younger objects to older objects.
4. most objects are never updated after creation.
5. older objects are updated less often than younger ones.
6. objects allocated together are likely to refer to each other.

7. objects allocated together are likely to die together.

These come in related groups, as shown.

I think most of us agree that (1) is the generational hypothesis. Dave
Mason proposes "persistence assumption" for (2). OK.

(3) is what started this discussion; Dave Mason proposes "ordering
assumption"; I would prefer "directional hypothesis" but fair enough.
It plainly comes in strong and weak varieties. (4) could be the
"immutability hypothesis". I'm not sure whether (5) is either true or
useful in most settings; do others have opinions on this? Anyway, it
could be called the "active youth hypothesis", or something.

(6) is surely the locality hypothesis, as Dave Mason has.

What about (7)? "pandemic hypothesis"?

All of these are suspect to [independent] measurement, of course.

Nick B