[ka-Map-users] evaluating ka-Map

Paul Spencer pspencer at dmsolutions.ca
Fri Sep 9 14:38:42 EDT 2005

I'll cover both of these in one response :)

There is no cost (or at least almost none) for toggling layers on the 
client side once a layer has been loaded.

The reason why I made the comment was because ka-Map's performance 
relies heavily on caching the tiles on the server.  Dynamic data (such 
as the type you are refering to) will require the cache to be deleted 
(or avoided altogether) on a regular basis, thereby negating most of 
ka-Map's benefit.  Additionally, the default configuration of ka-Map 
uses http headers to cache the tile images in the client's browser so 
that subsequent visits to the same spatial region/scale will grab the 
tiles out of the local cache rather than causing a hit on the server.

In the dynamic case, you would need to use much shorter cache expiries 
or use no caching on the client side to ensure that the most recent 
version is loaded.

Also in the dynamic case, assuming that a tile has been loaded and then 
the cache is cleared, scrolling would load the next tile from the newest 
data while the existing tile would have old data, which could 
potentially look really really bad.

In the case of WMS, Charlton is actually wrong (sorry).  ka-Map tiles 
are actually rendered by producing larger images and then slicing them 
up.  When a tile includes a WMS layer, it is rendered as part of a full 
map draw at the larger size and then sliced as part of the tile generation.

Hope this makes sense



Charlton Purvis wrote:
> Hey, Nate and Paul:
>>Just so I understand, is this because of the cost of loading all the tiles
>>when a layer is toggled on/off, or does it use some kind of optimization
>>that depends on static files?
> Nate, I've been playing around w/ ka-map lately, and it really is slick.
> Paul, please correct me if I'm wrong . . .
> If you're serving up a layer via ka-map that is really a WMS layer from
> another box, I *think* that ka-map will make n-WMS queries where n is the
> number of tiles in your given view.  In other words, I don't *think* that it
> makes one WMS query and then divides that image up into tiles.  If that is
> the case, then a dependence on a foreign WMS layer might be tricky
> performance-wise.
> The other question I have is:  Assuming that WMS is OK and that what I
> assumed above is way off base, what happens if the WMS layer is real time
> weather stuff like Nate is using?  Wouldn't it only query the WMS once to
> create the tiles = no real time update w/o doing something fancy on your end
> to either clear those from the cache on the hour or tag some kind of
> temporal index to each tile?
> Thanks, Paul, for such a dynamite product in such a short period of time.
> Charlton

|Paul Spencer                           pspencer at dmsolutions.ca   |
|Applications & Software Development                              |
|DM Solutions Group Inc                 http://www.dmsolutions.ca/|

More information about the ka-Map-users mailing list