[Openstreetmap] more distributed ideas

[Richard Fairhurst]

Before we go down the road of a complex distributed architecture, it's worth
thinking "is OSM running at 100% efficiency already?".

Pretty clearly it's not. Steve posted just a fortnight ago that he'd like to
replace mapserver with something faster
(http://bat.vr.ucl.ac.uk/pipermail/openstreetmap/2006-January/002088.html).
This really needs to be a priority.

As part of this, I'd suggest that we optimise heavily for delivery. In other
words: how can we precalculate certain data to avoid the server having to
crunch through the same old functions, time after time?

One example is to 'pre-project' data - in other words, rather than 
having to do
the Mercator calculations on lat/longs every single time, do the calculation
when the data is first loaded, and store the result in extra database fields.

Another is to store simplified segment paths. When you're fully zoomed 
in, sure,
show the full-point version. But when you're viewing a small scale map, you
actually only need a handful of points to generate the same line. Again, this
could either be done when the data is loaded/edited, or by means of a 
cron job.

Further in the future, a really good way to reduce server load is to use
client-side rendering technologies like SVG and Flash (for those who 
have them,
of course, not as the only option).

Apologies for being not as specific as I might like - I've only just started
playing with the source in the last few days.

cheers
Richard