A better way of phrasing those questions is "Is it worth it?" and "What is a good recognition method?". I'm not being picky, just trying to be helpful
Anyway,
for 3x3, COLL isn't really worth the effort.
For 4x4 and bigger, COLL is worth it, because EPLLs are 2-gen and easy on big cubes, and EPLL recognition is easier than PLL (particularly with parity).
For 2x2, COLL algorithms are way too long and slow, learn the 2x2 versions of each case from
http://www.speedsolving.com/wiki/index.php/CxLL
Rowe Hessler gives a good recognition tutorial, and a playlist of all his algorithms here:
http://www.youtube.com/watch?v=RHP2Hlgce-k&feature=channel_page
I've also got alternative CLLs to Rowe's on my youtube, and David Woner lists his, Rowe's + my CLLs on his site:
http://dtwoner.110mb.com/index.php?p=1_31_CLL
Though it seems to be a bit slow atm.
You can apply CLL recognition to COLL. There are plenty of threads about CLL elsewhere on this site if you search around a bit.