SRS - layer with two valid EPSG codes - conclusions

[Stefanie Weykam]

Thank you very much everybody and especially Ed McNierney.

Yes, this is it: "the grid itself actually moves on the Earth, so its
coordinates
in the new datum are the same as the coordinates in the old datum". 
And yes, I have been asked to provide the layer via WMS/WFS in at least one
more datum. Projection changes are not made in our application as such.

Duplicating or triplicating this layer wasn't my favourite option, since it
is actually the heart of the application and I had to change a lot more than
just that. 

I have come to the conclusion that probably the best way is to change the
web projection, reproject the other layers (few), assign the same projection
to the grid and provide it in just this projection, which will hopefully
suit everybody.
The alternative was to provide several projections and additionally comment
on the artefact in the layers metadata abstract (not sure if this is read)
and thus leave it to the clients to decide. I could also limit the display
MINSCALE, or perhaps this is even worse since it just hides the
error.Probably it is a mostly aesthetic question. The effect would be
visible only at small scales and only in case the client displays a grid
himself. Additionally, the error lies within the range of positional error
of our data which is not thought for small scale display anyway. It is not
likely either that somebody used the grid for any GIS operation.

But I don't fancy the idea to actually introduce a new error. I wouldn't
mind to hear the opinion of some of those who have more experience on
integrating data from different sources. Is this error acceptable for the
sake of more projections? If the data is displayed via a portal service,
would the user see my remarks?

I am also planning a new application which aims precisely on the
determination/detection of missing/incorrect projections of point locations.
In this case the grids (of various sizes) are not of interest for any other
cascaded servers. So, the other "lesson learned" is that I may just not
specify any grid layer projection at all - and thus assure that they always
fit to the underlying reference maps. 
stefanie