[mapserver-dev] Tile Access API

[Daniel Morissette]

Paul Ramsey wrote:
> I have opened an issue here:
> 
> http://trac.osgeo.org/mapserver/ticket/2581
> 
> ... and attached a working patch against trunk to implement the
> functionality.  I'd like to submit the attached RFC for vote and/or
> further discussion at the discretion of the Chair.
> 

Hi Paul,

Sorry if my comments come a bit late, I was on the road last week and 
saw the thread about this proposal only this morning.

<Note>
I'll start with the usual caveat: Please forgive my English, I'm French. 
;)  I spent last week at the Geo-evenement in Paris and after one week 
of forcing myself to speak "real" French I have a hard time expressing 
myself in English today!
</Note>

I'm not sure where to stand with respect to this RFC specifically... in 
my opinion, tiling is of little use without caching, and misbehaved 
clients or high traffic sites can easily bring a server to its knees. 
For this reason I tend to see this feature more as a toy than as a real 
feature... and since it comes with a risk for production servers (i.e. 
by encouraging use of non-cached tiling clients) I wonder if we should 
add this at all, or perhaps make it optional and disabled by default (so 
that an admin could prevent the use of his servers by tiling clients).

Of course, OpenLayers already does tiles-over-WMS and already has the 
potential of creating heavy load on WMS servers without the admin being 
able to do anything about it, but I was not involved in the decision of 
promoting use of non-cached-tiles-over-WMS by OpenLayers and if I had 
been asked I would have said that was a poor use of WMS.

In my opinion, tiling requires caching in most cases and there is 
TileCache and co for that, unless I'm mistaken it works for any WMS and 
not just MapServer... that should be the preferred solution.

Yeah, I know what you're thinking, TileCache requires Python and that's 
one more component to maintain when we already have MapServer installed.

This gets us to my second point... this RFC opens the door to caching 
tiles directly in MapServer and I'm not sure if we want MapServer to 
start managing a cache when it's real job is to produce maps.

Perhaps a solution if we want an easy to deploy tile service would be to 
re-implement TileCache in C and distribute it with MapServer... that 
would make just one more CGI binary to package with MapServer, but since 
that would be a separate program it could create tiles from any WMS 
service and not just MapServer.

I'm not completely opposed to the idea, but I'm not fully behind it 
either... am I the only one who is worried about the possible impact of 
this on servers?


Also, before we vote on the RFC, here are a couple of technicalities:

- Can you please add it to the RFC list on the website at 
http://mapserver.gis.umn.edu/development/rfc/ using the next available 
number (currently 43) and make sure you set the status of the Plone doc 
to "published". If your Plone account doesn't have edit privileges to 
that directory then let me know off-list and I can take care of that.

- Please commit and maintain a copy of the RFC in the mapserver/rfc 
directory in SVN

Thanks

Daniel
-- 
Daniel Morissette
http://www.mapgears.com/