============================================
 * Original From: Paul Edwards, 3:711/934.9
 * Original To  : Rod Speed, 3:711/934.2
 * Original Date: 1997-04-06 00:30
============================================

RS>> Pity it gives the WRONG result with the 9999 and 15001 parcels Paul.
RS>> That particular alg is useless, and we STILL havent decided at what
RS>> point we stop giving the spare share to the 9999 and start giving it
RS>> to the 150xx parcel as we move from 15001 to 15099.

PE>> Sorry, I meant compare (100-xx)/150xx to 1/9999.

RS> Where the fuck have you plucked that from ?

Well, that is how much of an ERROR I have produced, as a percentage of the
parcel being moved, in order to get the bonus share allocated somewhere.

It's quite logical.

PE>> You didn't reply to my second message on the topic either.

PE>> Which basically translates to:

PE>> (100-xx) compared to 1.

RS> God knows what this is about. Absolutely classic Paul Edwards cryptic.

Well, that's absolute number of shares being moved, not as a percentage,
but as an absolute value.

e.g., if xx above is 70.  We have to move 30 shares to get to a value where
the bonus share would be allocated, compared to 1.  The absolute value 1 is
smaller, so if the DESIRED OUTCOME is such that the absolute number of
moved shares is the least, then we compare 1 to 30 and choose 1.

On the other hand if the DESIRED OUTCOME is such that we wish to minimize
the size of the moved parcel, as a percentage of it's size, then we need to
compare 1/9999 with 30/15070 and see which one is smallest, and choose
that.

RS> I say again, until we decide what the we think the bonus share allocation
RS> should be in that test series of parcels, the rest is completely 
RS> irrelevant.

We are discussing the DESIRED OUTCOME, right?  I've listed two above. 
There are other DESIRED OUTCOMEs that I could list (all remarkably similar
to algorithms actually, but nevermind about that), e.g. the DESIRED OUTCOME
may be to minimize the movement of the parcels AFTER the initial
allocation, as a percentage of the package size, so in the above example,
we would actually be comparing 100/9900 to 100/15000, and choosing the
smallest one, to achieve the DESIRED OUTCOME of the minimum movement after
initial allocation.  I do not consider that to be a good DESIRED OUTCOME.

BFN.  Paul.
@EOT:

---