I will be implementing the OpenGL SC standard for an Intel graphics card, but the 1.0/1.0.1 SC standards state that the internal format for RGB must be using 3 bytes (assuming it means RGB8, with 8 bytes each texel).
The Intel cards only store textures in 1, 2, or 4 byte increments per texel. The RGB they support is RGB5 (R5G6B5). When you use RGB8, it is stored as RGBA8.
It is fairly common for cards to only support the RGB565 format internally so I was wondering if the standard was intentionally trying to go against the grain. Currently, I will have to list this as an area where I am not in compliance with the speciification, but I think this will be an area many will not be able to comply with due to hardware restrictions.
Also, the 1.0.1 standard still references table 3.1 for internal formats, but should really reference 3.2 since that table has been renumbered.