Normally drafts are restricted to a single block.
In sanctioned drafting (with sealed boosters), the distribution looks like this:
- Core set, the first set of a block, or a "big third set" like Rise of the Eldrazi or Avacyn Restored: three boosters from the same set, e.g. M12 / M12 / M12.
- The second set of a block: one booster from the new set, two from the first set (which is bigger), e.g. Dark Ascension / Innistrad / Innistrad.
- The (small) third set of a block: one booster from the third set, one from the secondset, one from the first set, e.g. New Phyrexia / Mirrodin Beseiged / Scars of Mirrodin.
(Note that, before Scars block, the pack order was reversed.)
As you can see from the list above, the size of the potential card pool will vary. M12 / M12 / M12 has around 250 possible cards. NP / MB / SOM is closer to 600. Thar larger card pool is trickier for players to learn (folks who play regularly will learn it incrementally, as new sets are introduced). I've heard of triple small-set drafts (e.g. Worldwake or Eventide), but many players say that these have very stilted environments -- both because the card mix is to small and because they were never designed to be drafted stand-alone, so a lot of the card pool assumes you'll have cards from the other set to round out your deck.
The main advantage of staying within a block is that block themes provide players with a reference point for making draft choices. Stepping outside of a single block presents players with additional challenges in evaluating cards against the "metagame" of the total card pool. This may be desirable if you're specifically looking to force players to think on their feet in an unfamiliar situation, but novice drafters have enough to worry about without the extra complexity.
Generally, the cards in a block are all designed to work together (in draft more so than anywhere else, actually -- only some of the cards in each set are intended to be "Constructed-worthy"). When you stray from that structure, you risk watering down the archetypes built into a set / block. For example, M12 has a lot of bloodthirst creatures in it. Red and black, the two colors with the highest concentration of bloodthirst, have access to common-level "enablers" to make bloodthirst decks run smoother: Goblin Fireslinger and Tormented Soul. Mixing up M12 and M11 cards would make it more difficult to draft bloodthirst due to the lower probability of finding these cards (although you might be able to compensate with M11 cards like Lightning Bolt and Prodigal Pyromancer). It's not to say that you can't break up blocks, but ideally you should have a pretty good reason for doing so. In particular, watch for sets with "parastic" mechanics like Kamigawa's soulshift -- diluting the card pool with cards from another set will greatly weaken these cards, making for a wonky and unbalanced drafting environment.
As I said in this answer, I don't think absolute pack-for-pack correctness is super important in a DIY draft, because I don't like going into pack 2 knowing you're absolutely "cut off" from any more set 3 if you didn't draft them in pack 1.