[OpenLayers-Users] maxResolution change

[Chris Holmes]

Tom Carden wrote:
> On 20/06/06, Chris Holmes <cholmes at openplans.org> wrote:
>> Schuyler Erle wrote:
>> > * On 19-Jun-2006 at  3:14PM EDT, Chris Holmes said:
>> >> Interesting.  Does it match it at the levels all the way zoomed 
>> out?  My
>> >> understanding is that Gmaps uses Mercator for the first five or so 
>> zoom
>> >> levels...
>> >
>> > Well, we haven't gotten that far yet. It's an interesting
>> > consideration if true. I'd had the impression they used Mercator all
>> > the way down.
>>
>> See:
>> http://cfis.savagexi.com/articles/2006/05/03/google-maps-deconstructed
>> for the real deal on what they're doing.
> 
> This doesn't seem to mention switching projections at all, it's just a
> thorough (and good) overview of how the tiles are defined and how to
> do mercator projection in javascript.
> 
> I recall that when Google introduced aerial and hybrid imagery for the
> UK there were alignment problems with the hybrid map due to projection
> issues. For example the outer circle of Regent's Park in London (so
> long the first thing seen on OpenStreetMap) was not circular.  That's
> corrected now of course.
> 
> Comparing with MS and Y! imagery (e.g.
> http://www.mapstraction.com/ljn.html using +/- keys on the Y! map to
> zoom) I don't see any significant differences in projection or
> alignment... only a pixel here and there.  This leads me to believe
> all three are using the same projection.
Hmmm...  But are the requests for the area to be viewed in lat/lon?  If 
all three are making the same requests in lat/lon I think they might 
look similar, even if ones in a bit of a different projection?  The 
thing to do would likely be to overlay them directly?  Though I'm not 
sure at all...  Would probably be better to overlay from a WMS that you 
know is mercator or lat/long and compare...  Any plans for mapstraction 
to support WMS as well?

> 
> Given the look of the maps - Greenland as big as Africa when zoomed
> out, circular things roughy circular when zoomed in - and class naming
> giveaways such as GMercator, I'd side with Schuyler and assume that
> it's turtles^H^H^H^H^H^H^H Mercator all the way down.  My own
> experiments with lining up canvas and SVG elements with Y! and MS maps
> would also suggest that it's Mercator all the way, since that's what I
> used and things seemed to line up.  I confess to not testing
> methodically or exhaustively though.
> 
>>
>> http://docs.codehaus.org/display/GEOSDOC/Google+Maps is what we've been
>> doing see:
>> http://dist.codehaus.org/geoserver/gmaps-geoserver_scripts/wms-gs-1_0_0.js 
>>
>> for the script that works with geoserver.  On close zoom levels it
>> switches to 4326
>>
> 
> Would be good to get this cleared up so people aren't making
> unecessary mistakes, and it looks like you looked into it in some
> detail.  Have you got an example of this up and running - perhaps away
> from the equator where differences would be most pronounced?
Sure.  I definitely don't know for sure either, and would love to get it 
all cleared up.  I've just had a few user reports and played a bit on my 
own.

For an example, see:
http://geo.openplans.org/cholmes/states1.html
and
http://geo.openplans.org/cholmes/states2.html

states2 is 41001 (mercator) the whole way, and it looks great fully 
zoomed out, but if you zoom in the states start to misalign.  Try 
florida or Maine, they start shifting to the left substantially.

states1 does EPSG:4326 the whole way, and you'll notice that when zoomed 
out it's way off, but when zoomed in its quite close, better than 
mercator.  Though if you get too close the inaccuracies of the dataset 
will show, but it doesn't have the big shift like mercator.

http://geo.openplans.org/cholmes/states.html shifts at zoom level 5 
between the two (though one up or down might align a bit better, it's 
still not perfect).

(let me know if the server goes down, I can kick it.  It's not a 
production server, just something I play with, so won't stay up a long 
time).


Chris

> 
> Best,
> 
> Tom.
> 
> !DSPAM:1003,44988130284412051017194!
> 

-- 
Chris Holmes
The Open Planning Project
http://topp.openplans.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cholmes.vcf
Type: text/x-vcard
Size: 269 bytes
Desc: not available
Url : http://mailman-viper.python-hosting.com/pipermail/users/attachments/20060620/67590fce/cholmes.vcf