The progress bar that doesn't convey any sense of progress (was Re: ...)

[Stephen J. Turnbull]

Brian de Alwis writes:
 > On 26-Nov-2009, at 9:31 PM, Martin Pool wrote:
 > > The idea that we should show total_bytes/total_time rather than
 > > instantaneous bytes/time is interesting to me.  I find it interesting
 > > to see it change over time.
 > 
 > 
 > I find the instantaneous rate useful.

The problem is that at least in my experience there is no "the"
instantaneous rate.  It varies from 0B/s to 500KB/s and back again
with a period typically 1-2s.

 > I sometimes have to interact with a slow, overloaded Subversion
 > server with bzr-svn.  Pushes will sometimes fail from a timeout the
 > network is slow (server is ~3000km from me) or if the server is
 > slow.  The instantaneous rates helps me in predicting whether the
 > push will succeed (lunch time!) or fail (stick around to try
 > again).

IME, the same information is available, more reliably, from the
average rates, but you need to interpret differently, which takes a
bit of time to learn.  You also have to watch the trend for a minute,
but that is well repaid by the higher reliability.

As long as the first statement is true for a non-negligible fraction
of users and the second is true for most users, average rates are
preferable.