It looks like Microsoft listened to the community and plans on switching latitude and longitude ordering for WKT and WKB:
http://blogs.msdn.com/isaac/archive/2007/12/27/latitude-longitude-ordering.aspx

This change will only be for WKT and WKB. That does make some sense, since the ordering is related to the SRID which is not included in the WKT and WKB, thus making it impossible to decipher without any prior knowledge. It would still be nice if the two geometry models shared a more common interface for anything that is remotely similar, but this is a good start.

I’m actually starting to like working with the geography type (to begin with I hated the idea of having two different although similar geometry models). It’s fun to do buffers on points and see that they don’t look circular on a projected map, because they are, well… projected. Showing the same circles on a 3D globe reveals that they are in fact perfectly circular. It makes distance queries so much easier than for instance using this approach.

One thing that still bugs me about the geography datatype though… why is the linear unit type that is used for calculating area, length, buffers etc. declared as part of the SRID (which in this case is always geographic) ? This requires one to always first have to look up the linear unit in the spatial_ref_sys table for each returned row (since homogeneous SRIDs aren’t enforced on the table) and divide by the factor returned to ensure that we know the unit we get back. In my mind a linear unit should never be related to a geographic coordinate system. After all, that’s why we only define linear units in projected and geocentric coordinate systems where they makes sense.

It would be nice if the methods either took a unit type as input (although that would break with the common interface I just wished for) or simply just always outputs the results in SI units (then it would just be a matter of multiplying with a constant well-known conversion factor later if you want to display it in for instance miles, thus separating code from presentation). Luckily most of Katmai’s SR definitions use meters as default, but you never know if someone chose to change the conversion factors in the system table so you will always have to check first.