[NNTP] 64-bit article counter extension strawman

[Andrew - Supernews]

>>>>> "Charles" == Charles Lindsey <chl at clerew.man.ac.uk> writes:

 Charles> And I think exactly the opposite. It is ridiculous using
 Charles> huge numbers (approaching 64 bits) to identify articles in a
 Charles> group where the maximum number of articles actually existing
 Charles> at any one time is unlikely to exceed 100,000, and probably
 Charles> nowhere near that.

If the number of existing articles isn't likely to exceed 100,000 then
non-cyclic 32-bit numbers will suffice anyway.

If exponential growth continues to the point where 32-bit numbers are
no longer enough, then you're looking at groups with perhaps 100 to
200 _million_ articles in at any given time (10 million per day, 10 to
20 days retention). Furthermore, five more years of the same growth
pattern and you're looking at groups with more than 2^31 articles in -
and so cyclic numbering breaks down.

Cyclic numbering would only be a win if you think that exponential
growth will continue long enough to break the 32-bit limit but then
suddenly stop or slow down. That seems improbable.

(63-bit numbering is safe for at least a further 40 years even
assuming that exponential growth continues unchecked.)

-- 
Andrew, Supernews
http://www.supernews.com