[Proj] What about datum shift via direct projection?

[Mikael Rittri]

Gerald wrote: 
> While there seems to be lull in the hot debate about separation of 
> church and state ... er ... datum and projection,
> [...] 
> Thus, why is it so necessary to bind the two operations so tightly as done
> in the proj.4 distribution? I cannot find a precedence for this concept. 

This post is not specifically about the PROJ.4 design (so I changed
the Subject line), but it is about how much datums and projections 
can and should be separated.  

There is method for datum shift that uses a direct projection.  

As an example, the old Swedish Grid is traditionally defined 
on the Swedish RT90 datum (ellipsoid: Bessel 1841) and using a 
Transverse Mercator projection with 
   central meridian: 15° 48' 29.8" E 
   scale factor:     1
   false easting:    1500000 m 
   false northing:   0 m
( http://www.lantmateriet.se/templates/LMV_Page.aspx?id=4766&lang=EN ) 

With this definition, one would need some datum shift method 
to transform between RT90 lon/lat and WGS84 lon/lat.

However, a simpler method, now recommended by the Swedish Land Survey
instead of a 7-parameter shift, is to start from the WGS84 datum, and than
tweak the projection parameters a little: just use a Transverse Mercator 
with 
   central meridian: 15° 48' 22.624306" E   
   scale factor:     1.00000561024
   false easting:    1500064.274 m
   false northing:   -667.711 m
( http://www.lantmateriet.se/templates/LMV_Page.aspx?id=5197&lang=EN ) 

A paper describing this technique is 
http://www.fig.net/pub/fig2006/papers/ps05_03/ps05_03_04_engberg_lilje_0670.pdf .

So, I have some rather vague questions to the readers of this list:
- What do you think of this technique?
- Is anyone else using it? 
- Doesn't the technique imply that a projected coordinate system 
  may have an ambiguous geographic coordinate system?  For the Swedish Grid,
  I can think of the geographic coordinate system as RT90 lon/lat, if I use 
  the traditional projection parameters. Or I can think of it as WGS84 lon/lat, 
  if I use the direct projection instead.
- If the correct answer to the previous question is "No, you fool", then what? 
  If I wanted to express the Swedish Grid, datum-shifted by the direct projection,
  in Well-Know Text, then I would be forced to say that the geographic coordinate 
  system is WGS84 lon/lat. But then the resulting CRS cannot be Swedish Grid, 
  because Swedish Grid has traditionally RT90 lon/lat as its geographic coordinate 
  system. 

I think direct projections for datum shifts are efficient and easy to
use, and normally as accurate as a 7-parameter shift.  But when I try 
to fit this method into the traditional framework that separates datum 
shifts and projections, and which insists that each projected CRS
has a unique geographic coordinate system, I run into problems.

Are these problems caused by inflexibility in the traditional framework? 
Or is the method of direct projection just weird? 
Or am I missing some good way to reconcile them? 
  
Best regards, 

--
Mikael Rittri
Carmenta AB
Box 11354
SE-404 28 Göteborg
Visitors: Sankt Eriksgatan 5
SWEDEN
mikael.rittri at carmenta.com
www.carmenta.com