Yup. You certainly could do it that way.
I am going with 3 vs 2 because I want a bigger divide in magic between arcane and divine. All arcane sub-classes will fall under the magic user class and all divine sub-classes will belong to the cleric class.
One reason for this type of organization is to limit multiclass sampling silliness. With only 3 actual classes in play, a given character can only have three classes: one type each of fighter, magic user, and cleric. It will still leave plenty of room for choice beccause each class will have a number of subclass options to choose from. From a rules/balance standpoint it saves a lot of headaches because the various abilities of all the subclasses will never be known by a single character.
For example, if a character starts out as a fighter and wants to multiclass then the options are either a type of magic user or cleric. Yet another style of fighter is not an option because that wouldn't be a different class.
This is why the number of actual separate classes needs to be low and I inquired about rogues in the first place.
Meh. I'd rather have a warrior/expert/spellcaster paragrim for three classes than fighting-man/magic-user/cleric. The latter was appropriately abandoned in Supplement I: Greyhawk and never looked back to.
There are more "thieves" in fiction than "clerics" (Aladdin, Robin Hood, Antilicus, Remington Steele, Grey Mouser, Han Solo, and more vs. Rolland, Friar Tuck, and..?) In addition, arcane and divine magic is a true D&Dism since most magic in fiction can heal, destroy, buff, and protect.
There are two questions in your query.
1.) Is a "rogue" archetype viable enough to stand on its own as a separate class, and
2.) What is the best way to implement it if so.
The answer to one IMHO is yes. Fighter (and almost every variant term; warrior, etc) has implied someone who is focused on combat first, other things second. A "rogue" implies someone who uses combat as a last resort, but uses skill, guile, and luck first to avoid combat. When that fails, he enters combat with less skill than a man who lives (and dies) by the sword. The "rogue" archetype could encompass a wide-variety of skilled characters; woodsmen/archers, scouts, con-men, gamblers, smugglers, acrobats, swashbucklers, etc.
D&D's thief (and rogue) class has always committed the cardinal sin of giving rogues a role, and then allowing them to suck at it. Too often, the nature of the skill system (or % system) meant low level skill characters failed far more often then they succeeded (akin to a fighter who could only hit typical monsters on a 15+). To add insult to injury, magical means of duplicating these skills were automatically successful, making them highly valuable. And at low levels where a rogue could contribute to combat (addding in occasional strong blows) the backstab rules were so confining as to make them worthless, and SA was so good they had to cripple it by making it only work some of the time.
IMHO, a good rogue class (system independent) should do the following.
1.) Have the widest (and best) chance of succeeding at skill-use. A rogue should be able to climb a wall, sneak down a hall, and listen at a door with a reasonable chance of success at low levels, his abilities to sneak, con, and such should rival magic at high-levels. (Bluffs so good they border on charm magic, hiding so good he's practically invisible) While everyone else can climb a cliff, I want the rogue waiting at the top, eating a sandwich, yelling "come on you guys, hurry up."
2.) A rogue shouldn't need to rely on magic (be it magical items or some innate talent with it) to do his job. Magic should enhance his abilities, not replace it.
3.) If cornered, a rogue should have abilities to misdirect, confuse, and outwit his foes with dexterity, panache, and bravado. Avoiding hazards with an almost sixth-sense like ability. While a fighter can anticipate a battle and a wizard knows a magical aura when he see's it, a rogue can sense trouble long ago and has focused his abilities to avoid or confront it.
4.) When all else fails, he's good at skirmishing combat. Standing in one place trading blows is a fighter's job, he should excel a landing occasional strong blows to unsuspecting foes (occasional being key), ducking and weaving in combat to strike at foes, and mostly picking off stragglers, wizards, and other "squishies". Against an goblin or guard, he can hold his own a bit. Against an ogre or dragon, he better run behind the fighter!
All versions of D&D rogues promise this, most fail in one regard or another. Mostly, they insist on letting magic trump mundane skill and try to assemble glass-cannon combatants (I'd gladly train 1/2 my SA dice for a d8 HD and 3/4 Bab in 3.5 for example). The problem is exactly what you point out; rogues (in 3.5) ended up Trapfinding melee-fireball attacks, and in previous D&D they were walking lockpicks.
If your system makes a "fighter" that can do everything I just outlined above, I'll gladly accept your premise. Otherwise, I want my rogue class!