I am gathering statistics on how many image sizes would be affected by changing upright to use square bounding boxes, and optionally also changing the default scale factor to 1.0 at the same time (assuming the default scale factor was mostly used for 3:4 portrait aspect-ratio images). The statistics-gathering code can be found at https://github.com/cscott/wp-image-info

For frwiki, there are 1,242,985 references to non-commons images, of which 17,870 use the 'upright' specifier. If upright were changed to use a square bounding box, 1049 would change size by more than 5px if the default scale factor was left at 0.75, and 1053 would change size by more than 5px if the default scale factor was changed to 1.00.

On dewiki, there are 1,786,779 references to non-commons images, of which 38,624 use the 'upright' specifier. If upright were changed to use a square bounding box, 1963 would change size by more than 5px if the default scale factor was left at 0.75, and 1764 would change size by more than 5px if the default scale factor was changed to 1.00.

Statistics collection on enwiki is still ongoing.

(In reply to C. Scott Ananian from comment #1)

Statistics collection on enwiki is still ongoing.

Any news?

For enwiki, there are 4,949,488 references to non-commons images, of which 37,185 use the 'upright' specifier. If upright were changed to use a square bounding box, 1971 would change size by more than 5px if the default scale factor was left at 0.75, and 1866 would change size by more than 5px if the default scale factor was changed to 1.00.

So roughly 0.75% of the images on a wiki use upright, and 0.04% would change size if 'upright' used a square bounding box.

Given that the # of size changes are roughly comparable for the 0.75 and 1.00 default scale factor, I'd suggest that we should use the 1.00 scale factor, which will remove one more magical constant from wikitext semantics.