scheduled transactions

Rob Browning rlb@cs.utexas.edu
09 Feb 2001 00:38:47 -0600


Mike Talbot <mike.talbot@zepler.org> writes:

> I think this would create a much more flexable system: a number of
> 'standard' frequency functions can be defined (along the lines of
> Joshua's FrequencyEnum), but there is then scope for a 'user
> defined' function(s) (presumably written in scheme) so users can
> create the most complicated scheduled transactions they can dream
> of!

It's been a while since I thought much about scheduled transactions,
though I had been meaning to do them myself.  (Of course, if someone
else gets to them first, then that's great too.)

In any case, though I thought about this as a possibility, if I
understand your correctly, this approach, while very flexible, has the
distinct disadvantage that you can't optimize any forecasts about the
future (including simple calendar display).  The only way to find out
when things are going to happen is to probe every possible event time.
And since I had also imagined that we might want to be able to
schedule things to the hour or to the minute, this could quickly
become very expensive.

-- 
Rob Browning <rlb@cs.utexas.edu> PGP=E80E0D04F521A094 532B97F5D64E3930