Consider revisions to IIIF Image spec to disallow calculated size of zero pixels

[beaudet]

Please see ProjectMirador/mirador2#3 which describes a specific use case / Mirador bug that led to a custom IIP patch to prevent a 400 error response when the computed size of an image by IIP ends up being zero. I submit that it's better to apply a minimum bound of 1 pixel to such calculations. To summarize: if the original request of an image with w:h aspect ratio of 2 is requested with a one dimension size constraint of "1," then the calculated height rounds to zero. That isn't a valid image size, yet returning an image that's 1x1 pixel in such cases seems perfectly reasonable as the aspect ratio of images resized to small integers can be substantially different from the original aspect ratio anyway. Returning a 400 error in response to a valid client size request seems to just complicate client logic. It would be better in my opinion to tweak the language in the spec to say the server "shall" apply a minimum bound of 1 pixel to calculated image sizes rather than returning a 400 bad request error.

[azaroth42]

As a backwards incompatible change, this could only be changed in Image 4.0.