It looks like a Schmidt Roof prism will provide an erect image at 45 degrees and a ABBE Roof prism an erect image.
Both will create a longer light path in a compact optical block, long enough to be picked up by a macro/relay lens or even to come within the range of the camera inbuilt lens 'macro' setting without the need of a magnifying diopter (less chances to add unwanted abberations).
Also by picking the GG image from such a distance (~230mm +), IMO, will be beneficial in regards to the hot spot problem.

Quite right about the Schmidt and Abbe roof prisms. So far I can only find a schmidt of an apropriate size. I'm tempted just to go with one of these but I really don't want to have to deal with the whole 45 degree angle thing. I'd prefer to keep it linear.

I've been thinking about replicating the system that most high end binoculars use: two porro prisms. This seems to provide the better light transmittance and contrast as the beam is only re-directed twice. Should be interesting to experiment with at least. I just have to find a place to buy these things :(. Much easier said than done! I'd of course prefer to use common parts rather have custom optics made.

Frustrating to say the least.

I've tried and still use the two x 90 degree. prisms. If you are happy with the 24mm x 18mm movie frame, 2 x 40mm x 40mm x56mmm x 40mm across all faces will work via a Century Optics 7+ achromatic dioptre.

With 7+, the path is a bit tight in that the gg disk has to run beneath the front (rear-facing) prism which rules out using a condenser unless you can flatten one edge of it. A 5+ in the path might be more managable.

True binocular style Porro prisms may enable you to make the thing a bit smaller and lighter but you are still stuck with working to the inside faces of that ninety degree angle and if the outer corners have been rounded to provide the circular fit, then there may be wasted glass on the hypotenuse face you can't use to scan the rectangular image.

Just posted this in another thread but felt is was quite applicable here:

Anyway, something just occured to me... something that I should have thought of a LONG time ago with regard to flipping - and honestly I'm surprised it has never been used yet by the mini35 guys etc. It sure beats using a prism.

If you had the right focal length lenses you could do this (something I think I'm going to try with my homemade adapter):

http://www.weet.us/idea.JPG

It's so incredibly obvious... I can't believe I didn't think of this before. I'm not sure it would correct the left to right issue (would have to think about that some more) but it would solve the upside down problem.

*feeling like a complete moron*

Just thought about this some more and I am quite sure it would flip the image every direction we need.

Honestly, don't know why I didn't think of this before. What do you guys think? Am I going crazy?

wouldn't such a lens get terribly out of focus?

Not if done correctly Ocar. A similar setup is used, for instance, in submarine periscopes all the time to relay the image long distances (ie from outside the sub to the users eye). It's a very, very simple optical phenomena. I guess that just proves that the designers of the mini35, movietube, and I are not professional optical designers!

Matt Ziegeler asks in another thread:
get a lens like that in the diagram?

The shape of the lens itself isn't really the important part. The important aspect is having the correct focal length. It may be possible to do this with stock optics. I'll have to look. I suspect it is but it may require a custom piece of glass.

Oddly enough...I had sorta thought of this...but since I have so little knowledge of optics I assumed that someone already knew why it wouldn't work.

I'd assume it has quite a few negative aspects....increasing the size of your adaptor, losing more light, and it probably has to be custom designed so there is no distortion of the image. You might even need another lens to act as a second fresnel.