Google maps w/ satellite imagery

[Arnulf Christl]

Charlton Purvis wrote:
>>>So it sounds like you are generating all of the tiles statically and
>>>then just serving those up via javascript?  Has anyone tried using
>>>mapserver to tile on the fly, i.e. use mapserver to generate image tiles
>>>at certain zoom levels based on feedback of the user searches.  Then
>>>store tiles in a server cache.  When users request a map, have the
>>>server check the server tile cache, before kicking off mapserver to
>>>create new tiles.
>
>
> Believe you me, I'd love to do this.  But what prevents me from doing this
> and w/o the help of 10,000 CPUs to burn like google maps, is the fact that
> my underlying data changes constantly.  I'm in the near real time ocean
> observation business, and I try to cache as much as I can.  Considering
> that I go out around the clock fetching and aggregating ocean
> observations, an image that a user requests could be saved, yes, for
> caching purposes, but once the underlying data changes, that cached image
> is SOL.  The caching that I do is hourly to keep the initial maps and the
> common extents performing somewhat snappily.  But these hourly caches have
> to be refreshed whenever any new data comes in.
>
> While the google maps stuff is sexy, what would really blow me away would
> be a site that is that fast AND is also based on constantly changing
> underlying data.
>
> Charlton
>

grumble... but this is what MapServer gives us ervery day!? WE are not
as fast because WE do not have 10,000 CPUs and level 1 fibreoptix
internet nodes - thats all.

Warning: PointOfView!
If you think this is too off topic just tell me i'll quit it.

Maybe we should also discuss this topic the other way round. Why not
tell Google to rather stick to standards and let them 10,000 CPUs burn
MapServer images?
:-)
Just because Google is a bigshot in the web does not *have to* mean that
they are stupid - ahhm - wont listen to you experts.

I bet that Google could operate a bunch of MapServers just as fast as
those cached and tiled googlemaps we see now. And they would not be
restricted to scale levels. I also repeat myself when i say that google
maps only *seem* to be faster psychologically - they definitely are not
faster *physically*. Performance at this stage is eaten up by opening
and closing requests in the browser. Believe me, we have this problem
every day when we try to loose weight with our client interfaces. The
WMS images are definitely *not* the bottleneck.

Googlemaps is a nice experiment but we KNOW that it is no good in the
long run. Have a look at the scales and resolution satellite image they
offer, this is a minuscule fraction of all satellite and ortho images
available world wide - as WMS. And googlemaps will never ever be able to
be integrated with any of those datasets. What a waste of energy.

Hey - we are falling back into medieval raster tile viewers - i thought
that we had overcome those times long ago. Somebody should really point
this out to Google, else they might make a big mess of the standardized
infrastructures that are slowly emerging all over the place. We should
not let this happen.

We have MapServer installations running with TBs of aerial photography
data which never take longer than a few millisec to answer. You simply
cannot make this any faster (i know, i know Frank and Steve and Daniel
and Yasseffa and all you other cracks *do* get it faster every now and
then...).

Our private little problem rather is the client. It is getting way to
heavy because people cry for all those little gadgets functions and
tools. Now it has to be animated panning and we just finished animated
tabs...

:-)

Best, Arnulf.

--
--------------------------
Arnulf B. Christl
--------------------------
Mapbender User Conference:
http://wms1.ccgis.de/ewiki
--------------------------
http://www.ccgis.org
http://www.mapbender.org
--------------------------