[OpenLayers-Users] TileCache and KaMaps

[Christopher Schmidt]

On Tue, Dec 05, 2006 at 03:55:28PM -0600, Tim Langlois wrote:
> I know this was asked a while back by someone, but I could not find an
> answer. What are the advantages/disadvantages to using TileCache vs.
> KaMaps?  I am currently using KaMaps (via OpenLayers) to tile and cache
> mapserver images on the backend.  It works well.  I wasn't sure if
> moving to TileCache would improve performance and/or tile quality.  Has
> anyone tried both?

TileCache was originally designed as a simple tile caching server. It
didn't create larger images and split images up, one of the things that
ka-Map does to work around limitations in MapServer. 

In 1.2, metaTiling support does exist in TileCache. It's not extremely
well tested yet. It also doesn't hook directly into mapscript (and
there aren't plans to, though it could in the future) to use added
capabilities of MapServer like ka-map does. It also (due to lack of
support in the Python Imaging library) doesn't support chopping up
transparent tiles and maintaining transparency: this means that you
can't metaTile transparent images.

TileCache is young. It hasn't been tested on dozens or hundreds of
installs like ka-Map has been.

TileCache is written in Python. It doesn't require mapscript. It can be
run on a service which is far from your original service (via WMS
cascading). The code is clean. It runs under CGI or mod_python.  

TileCache supports multiple request formats. WorldWind, WMS-C, TMS.

TileCache supports multiple data backends: mapserver (via mapfiles), WMS
cascading, mapnik, and is easy to extend to anything else -- for
example, one of the next additions I'm going to look into is an ArcXML
layer. This means that if you have a custom data source, it's probably
easy for you to wrap it in TileCache. 


TileCache is, for the most part, a replacement for ka-Map. However, it's
an immature project, and may not fit all the needs that everyone has.
The metaTiling support, in particular, is not extremely well tested, and
that is the primary benefit that ka-Map offers. However, based on
empirical testing, I feel that it's likely that for many cases,
TileCache will be much faster than ka-Map, and I think that it's more
likely to be maintained and extended in the future than ka-Map's
rendering tool.

Personally, I'm working to migrate all my ka-Map based instances to
TileCache. If TileCache doesn't do it, then my steps will be to make 
TileCache do it, rather than setting up ka-Map. However, I have not yet
migrated any projects which require metaTiling to TileCache, so I'm not
yet convinced of its stability at that particular task. Anything which
doesn't require metaTile support, I would absolutely recommend
TileCache: I'm not to the same point with anything which requires
metaTile support yet.  

Regards,
-- 
Christopher Schmidt
MetaCarta