[OSM-dev] Creating Easy Map Embedding

[David Earl]

Why don't we create a distributed tile server, so the load can be spread 
around? Then we do have "infinite resources like Google".

It could be done geographically, but that might be uneven - for example 
consider what would have happened to the poor person who was hosting the 
Oxford area today.

Would an index be too big to distribute? If we worked on zoom 12 tiles, 
you'd want say 100 bytes for mapping each tile index to the relevant 
server (which I'd allocate at random and shuffle around as tiles are 
updated) the index would be about 160Mb: big, but not enormous. Equally, 
we could, like the DNS system, have several index servers, but fewer 
than the number of tile servers. This would allow us to distribute tiles 
at finder granularity as well with bigger index files.

So a tile request would immediately and be bounced to a random index 
server, which would bounce to the host for that tile, or return the URL 
to the client who can then make the requests direct to the myriad tile 
servers. (We could batch up the index requests in that 2 step version as 
someone is likely to want more than one neighbouring tile).

David

On 22/10/2007 14:43, Christopher Schmidt wrote:
> (This is a summary of a discussion that was had on IRC earlier today.
> Much of this content has also been put into a wiki page at
> http://wiki.openstreetmap.org/index.php/Creating_Easy_Map_Embedding .)
> 
> Google Maps offers an iframe based embedding solution. It seems like 
> adding something similar to OSM might increase adoption, or at least
> awareness, of the project, but there are a number of social and
> technical hurdles that would block such adoption.
...