On Sat, 26 Aug 2006, Richard J. wrote:

John Rowland wrote:

Better still, I know the walk from my house to the nearest train
station takes 12-15 minutes, but the JP suggests a 20 minute walk
to a nondescript point on a hail and ride route which is further
than the station, and then getting the bus back to the train
station.

The default walking limit is 20 minutes at average speed. If it deems the
station is further than this, it may not offer an all-walking route.

But, according to John, it *does* then offer you a bus route *which
involves walking further than to the station*. My guess is that the logic
looks like this:

TO FIND ROUTE FROM A TO B:
LET T = TIME TAKEN FOR CARBON-BASED BIPED UNIT TO WALK FROM A TO B
IF T AS FAR AS FAT/LAZY/CRIPPLED C. B. B. U.S WILL WALK:
RETURN WALK FROM A TO B
ELSE:
LET C = NEAREST BUS STOP OR WHATEVER TO A
RETURN WALK FROM A TO C, TAKE BUS FROM C TO B

Which will merrily emit two-leg routes in which the walk is longer than
the direct option.

This is slightly surprising; i would have thought the algorithm would be
something more general (Dijkstra's algorithm or A* or similar), which
would not make this mistake.

tom

--
Well we know where we're going, but we don't know where we've been