Back in the Good Old Days, multiclassing in D&D was complicated, unbalanced, and overpowered. Mind you, we didn't know that then. Then, multiclassing was the greatest thing since sliced bread. To the true aficionado, it was obvious that you wanted to play a demi-human (as elves, dwarves, etc. were called in those far-off times) so you could multiclass, at least until the level limits started to kick in. When Unearthed Arcana came out (and later, 2nd Edition) and raised the level limits, it got even better. True, you were a level or so behind your single-classed human counterparts, at least until you got past 10th or so (although the different experience tables for different classes meant that you could metagame around this to some extent), at which point you began to lag further (but you were pretty near the level limits anyway, and lots of campaigns never got far past that point). But if you looked at the abilities of, say, a 10th level fighter or magic-user and put them up against the abilities of a 9th/9th level fighter/magic-user, the choice was a no-brainer.
Then D&D 3rd Edition came along and changed all that. There was a good deal of vociferous objection to the new multiclassing system on the various D&D-related online message boards and discussion lists before 3E came out, mainly on the grounds that it was going to make multi-classed characters too weak (and truth be told, I was one of the objectors, so I understand the attractions of the old system). Once the new edition had been in play for a while, however, it became clear that the new multiclassing system really was a substantial improvement over the old one: it was simple and straightforward, and it neatly balanced the tradeoffs of versatility vs. specialization. Except for one particular area: spellcasters. The general opinion seemed to be that D&D 3E hosed multi-classed spellcasters, and so when the "3.5" edition came out, a bunch of new prestige classes were added to "fix" this problem.
Now don't get me wrong: I like spellcasters (and their relatives, psionics, to whom many of the same considerations apply). In fact, I believe that, in the long run, spellcasters are the most powerful classes in the game--which is why, although I agree that in some respects, multiclassing spellcasters in 3E needs to be improved (and I'll discuss the major such improvement that I favor below), I don't agree that they need the drastic "improvements" of D&D 3.5. The problem is that these "improvements" go way too far; they make multiclassing spellcasters almost as unbalanced as it was back in the Good Old Days, which as we have seen, weren't really so good from a balance standpoint. I used the example of a fighter/magic-user under the old editions above, but if you took a 10th level cleric or 10th level magic-user vs. a 9th/9th level cleric/magic-user, the choice was even more of a no-brainer. With that in mind, let's take a look at the poster child for the 3.5 version of multiclassing spellcasters: the Mystic Theurge.
Here's the comparison that's usually touted: a cleric 10/wizard 10 vs. a wizard 20 or a cleric 20. The cleric 10/wizard 10, it's said, will basically be a turnip in a 20th level party because (a) his spell DC's are too low (because he only has access to 5th level spells instead of up to 9th) to make the monsters' saves a challenge; (b) his caster level is too low to allow him to penetrate any 20th level monster's spell resistance. So to "help" this poor multi-classed turnip, we instead make him a cleric 5/wizard 5/Mystic Theurge 10. Now he's a respectable member of the party; he's got 8th level spells, which are much closer to 9th, and he's a 15th-level caster, which is much closer to 20th.
But (goes the criticism that's usually touted) the Mystic Theurge now has 8th level spells and is a 15th-level caster in not one but two classes. From a spellcasting point of view, he's a cleric 15/wizard 15, which is the equivalent of a 30th level character, and the only things he's giving up are (a) 10 effective levels of cleric for purposes of turning undead; (b) two wizard bonus feats. Sound like a fair trade to you? Evidently it did to Wizards. (And by the way, the imbalance is even more glaring if you consider, say, a cleric 3/wizard 7/Mystic Theurge 10 vs. a cleric 3/wizard 17; the cleric/wizard is almost the equivalent of a wizard 20 anyway, with only three levels of cleric, but being a Mystic Theurge gives him effectively thirteen levels of cleric for spellcasting purposes, so now he has not only 9th level wizard spells, but 7th level cleric spells, and sacrifices practically nothing. This example makes it obvious that practically every spellcaster will, or should, take the Mystic Theurge class, and last I checked, that was supposed to be a red flag that a class is unbalanced. Go figure.)
Now my position should be obvious from the title of this rant and from the previous paragraph. I don't care how you slice it: no one will ever convince me that a weaker turn undead and a couple of bonus feats are a fair trade for ten extra levels of spellcasting power. Anyone who thinks this is balanced is, in my opinion, trying to have their cake and eat it too. Back in the Good Old Days, this would have fit in with the rest of the system, but to revert to that mentality is to throw out the great insight that 3E multiclassing brought: that the class level is the right basic unit of measure for assessing and balancing tradeoffs when building characters. To get the abilities of two class levels for the price of one is obviously unbalanced once you have this insight, and viewed in this light it is equally obvious that what you give up for each level you take of Mystic Theurge is worth far less than a class level, so it can't possibly balance what you gain. (As another by the way, the other prestige classes on the same theme in the DMG, though not always "pure" spellcasting classes, all suffer from the same disease; it's just not always as glaringly obvious because there aren't two spellcasting classes involved. But if you look at, say, the Eldritch Knight, and realize that it's basically gaining a level of fighter and a level of wizard, minus far less than a level's worth of tradeoffs, you'll see what I mean.)
But then what do we do about the poor turnip of a multiclassed spellcaster? First of all, to think of him as a turnip misses the point of multiclassing. Multiclassing is about versatility. True, the cleric 10/wizard 10 can't get past the spell resistance of the pit fiend, but if he's being played right, that's not what he's there for. Taking out the enemy big guns is the job of your big guns--the wizard 20, in the case of spellcasters. The cleric 10/wizard 10 shouldn't be trying to penetrate the pit fiend's spell resistance; he should be casting buff spells on himself or other party members, casting prayer and so forth to help the party in combat, and so on. In other words, since he's not so good at achieving high save DC's or penetrating spell resistance, he should be using spells that don't require saves or penetrating spell resistance. To want the versatility of multiclassing and the power of those big flashy spells is, as I said, to want to have your cake and eat it too.
But lest I be accused of attacking a straw man, let me also admit that, as I said earlier, there are ways in which multiclassing in 3E can legitimately be improved. The major such improvement that I agree is necessary is the following. As is often noted, non-spellcasting abilities, for the most part, stack easily when you multiclass. A fighter who multiclasses to, say, rogue still gets some benefit to his attacks and saves from those rogue levels. A druid/ranger stacks her ranks in Wilderness Lore (now Survival) from both classes. Why isn't there a corresponding way for spellcasting abilities, or at least some of them, to stack when multiclassing? This is really the root cause of all the complaints, and there ought to be some way of addressing it without going overboard as the Mystic Theurge does.
It should be noted that not everyone sees even this as an issue; it can be argued, and has, that casting arcane spells and casting divine spells are so different that to expect them to stack the way that base attack bonus stacks between a fighter and a rogue isn't logical. I agree with this position to some extent, but not completely. There is an obvious intermediate solution, caster level stacking (with some limitations), and this is, I think, the right one; my own take on it will be published in an upcoming sourcebook by PT Games, Inc. (you can read about it here if you're interested), and no doubt there are plenty of others along similar lines. But such solutions are tailored to address one genuine, specific issue with multiclassing spellcasters, without breaking the whole system.
So to summarize: in earlier editions of D&D, multiclassing was too good. 3rd Edition introduced a much better way to handle it and keep it balanced, but it made multiclassed spellcasters a little too weak. But the 3.5 solution, the Mystic Theurge and its ilk, goes too far back in the other direction. Spellcasters deserve a rules mechanic that will help some of their abilities stack with multiclassing, similar to the way attack and save bonuses do; but to give them two levels' worth of full spell ability for the price of one, even with a few small tradeoffs, is way unbalanced. There are better ways of addressing the problem without going back to the Good Old Days.