On Fri, 25 Apr 2008, Gunnar Wolf wrote:
> Diffs are often quite small. The time needed to establish a
> connection is often comparable to the time a diff takes to be
> downloaded... So, even with pipelining and all, storing too many
> diffs becomes irrelevant.
There's also the (non-trivial) overhead of actually applying the diff
to consider. [The equation for the break-even point is something like:
package.gz_size/slowest_speed =
(num_pdifs)[ave_pdif_size/slowest_speed+ave_pdif_apply_time]
num_pdifs = package.gz_size/link_speed
------------------------------------------------
[ave_pdif_size/link_speed+ave_pdif_apply_time]
where link_speed is the slowest link speed we think is reasonable (56K
modem?) ]
I know for most of my machines, the number of pdifs at the break-even
point tends to be around 20. [Super fast links, but not-so-hot cpus.]
Don Armstrong
--
The state must declare the child to be the most precious treasure of
the people. As long as the government is perceived as working for the
benefit of the children, the people will happily endure almost any
curtailment of liberty and almost any deprivation.
-- Adolf Hitler _Mein Kampf_ p403
http://www.donarmstrong.com http://rzlab.ucr.edu
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
