[Mapserver-dev] EPSG:NONE

[Daniel Morissette]

Jan Hartmann wrote:
> Daniel Morissette wrote:
> 
>>
>> Perhaps MapServer should try to support SRS=NONE, but we need to think 
>> about the way to support this in a mapfile and internally.  For 
>> instance the spec also states that
>>  "Clients should not attempt to overlay information whose SRS=none
>>   with other information."
>> This restriction is important otherwise the behavior is undefined if 
>> you request a geographical layer using SRS=NONE or if you try to 
>> overlay two layers that don't have a SRS.
>>
> 
> Daniel, what is the problem with overlaying two layers without an SRS? I 
> am working with unreferenced historical maps with superimposed 
> architectural drawings digitized from the maps.  I would like to publish
> these as WMS maps in a few months, but for now everything needs to be 
> geo-referenced. This gives some distortion as the maps are not exactly 
> to scale. MapServer can overlay non-referenced images without problem, 
> so why could its WMS version not do the same?
> 

If you know that the two layers with SRS=NONE represent the same area 
and can overlap then there is no problem (which seems to be the case 
with your historical maps).  But just putting SRS=NONE in a capabilities 
document doesn't tell the client that it is safe to combine two layers. 
  OTOH when two layers have SRS=EPSG:4326 then it is clear that it's OK 
to draw them on top of each other.

We could let MapServer overlay images with SRS=NONE and let the users 
deal with the unpredictable results I guess.

BTW, should we allow zooming on layers with SRS=NONE?  Does this assume 
that the BBOX coordinates are in pixels in this case?  I think this 
behavior is clarified with the CRS:1 that replaces SRS=NONE in WMS 1.2, 
I would have to re-read that part of the spec before we implement anything.


> As to overlaying two layers with different SRS's, I'm not sure if the 
> results are "undefined". You can perfectly well combine coordinates from 
> different base systems; it just doesn't make sense, as you get maps with 
> wildly divergent parts. That is what people do who try to display a 
> latlong map in a geographic projection and "don't see anything" as a 
> result. It's not impossible, you just don't get what you expected. I 
> guess that is why the spec says you should not attempt it: you can do it 
>  but it is not wise.
> 

MapServer is smart enough that if you try to overlay *WMS* layers with 
different SRS then it will reproject them to use the same map projection 
and the result is fully valid.  That's different from having two layers 
with SRS=NONE since NONE has no absolute reference to anything.


> As I said, I would like very much this SRS=NONE capability. I refrained 
> from asking for it, since I thought it was outside the WMS specs, but 
> now I know it is in 1.1.1 (thanks Tom), I would like to add  my €0.02 
> request for it.
> 

If we implement this as stated in the spec then your problem won't be 
solved since we won't allow overlaying layers with SRS=NONE.  I agree 
personally with this restriction of the spec, but I understand why you 
need this and I guess it would make sense to leave the implementation 
flexible enough to let MapServer overlay layers with SRS=NONE and let 
the user deal with whatever result they get.

Daniel
-- 
------------------------------------------------------------
  Daniel Morissette               morissette at dmsolutions.ca
  DM Solutions Group              http://www.dmsolutions.ca/
------------------------------------------------------------