I have been using V4, but I recently read that Andrew said there is not currently any advantage to using V4 profiles for anything (if I understood him correctly)...
thanks,
v4 profiles are better for sure in the non-practical sense, however not all programs support them so in practice they could be potentially infinitely worse, it depends and sometimes mixing v4 and v2 can lead to odd things if the programs don't handle everything perfectly.
v2 leaves lots of things up in the air, such as the exact chromatic adaption method used, was it actually used or not, what was the assumed perceptual intent PCS, what sort of black point compensation assumption and so on that could trick smart CMM since they have to guess what the real deal is while v4 makes things more explicit and some of the issues may have somewhat held back larger scale adoption of perceptual intents and smart CMMs.
v4 also gives more bits of precision to certain things such as gamma exponents and such, I believe.
I really think that it's more than time that programs started all supporting v4 and slowly doing away with v2.
v2 specs were really something of a mess. It seems a shame that v4 still get shoved to the side all the time and called worse, etc. when it is v2 that should be getting ragged on and every thing should be receiving pressure to switch to v4.
v2 specs are actually open to far more potential issues and mistakes and a lot more weird stuff like pseduo colorimetric+perceptual called colorimetric and on and on.