from @jwalterlarson
Currently the Accumulator works on the assumption of uniform step size in the
accumulation cycle. This has the advantage of supporting a large share of what
is out there. Furthermore, it does this without introducing extra floating
point uncertainty (i.e. dt) in the result.
But, we should start confronting the idea that the averaging cycle might be best
defined in terms of a floating point value (like an hour) and also that we may
wish to start supporting nonumiform step sizes
(copied from MCT's old bugzilla)
from @jwalterlarson Currently the Accumulator works on the assumption of uniform step size in the accumulation cycle. This has the advantage of supporting a large share of what is out there. Furthermore, it does this without introducing extra floating point uncertainty (i.e. dt) in the result.
But, we should start confronting the idea that the averaging cycle might be best defined in terms of a floating point value (like an hour) and also that we may wish to start supporting nonumiform step sizes