Strange difference in LAB -> AppleRGB conversion between colormath and Photoshop

[jan-warchol]

@gtaylor this might interest you.

When I switched from Photoshop to python.colormath as the tool for doing color space conversions (see commit 5189cb9f0d092c456b73bdf2a6a4abd2c0fdf6e3), I noticed one strange discrepancy in conversion results: Photoshop converts Lab(63, 66, 44) to #f63e3d (in AppleRGB) while python.colormath yields #ff3e3d.

What's most surprising, when I converted Lab(55, 66, 44) - i.e. the same hue, just 8 units dimmer - to AppleRGB using both tools, the results were almost the same: #e0272e in Photoshop vs #e1272e in python.colormath. The coordinates of all other colors I converted were also almost the same (I've only observed some differences of 1 unit which can be attributed to rounding errors).

Who is doing the conversion right, and who is wrong?