[OSM-dev] Slippy map way forward -- OpenLayers and/or Mapstraction?

[Christopher Schmidt]

On Sat, Jul 01, 2006 at 06:59:14PM +0200, Lars Aronsson wrote:
> Christopher Schmidt wrote:
> 
> > > You mean as in setting a 48h timeout, that's easy to do in Squid. But
> > > can you check if a BBOX has been updated?
> > 
> > No. This is one of the long-term existing problems with OSM tiles in
> > general though, so this isn't new to using a 'real WMS':
> 
> Wikipedia uses Squids (in USA, Korea, and the Netherlands) for 
> readers who are not logged in.  But users having the login cookie 
> set, always drink fresh water directly from the live PHP servers.  
> 
> Could OSM do something similar?  I.e. serve non-logged in users 
> from a week-old cache, but allow logged in users to access more 
> fresh data?  And would it make any difference for us?  Does OSM 
> have lots of non-logged in users?  Do we know the proportions?

I think that this is probably not the best course for optimization --
that some further work needs to be done on improving tile rendering
speed before we start sending lots of stuff uncached, even to logged in
users, since a single logged in user can certainly cause a significant
performance hit. There's also the fact that without caching right now,
the site is mostly unusable for 'browsing' --  as you said, 3 seconds
per tile is 30-45 seconds per zoom level, so if this data was generated
live...

In any case, I think there are probably other optimizations mentioned in
this thread which would be more worthwhile to pursue than this one,
although obviously I could be wrong.

-- 
Christopher Schmidt
Web Developer