What is a Class?

In traditional Dungeons and Dragons, the first physical step of character creation is to roll some dice to determine your new character's ability scores (there is often a step 0 before this which is envisioning the type of character you might want, but I'm omitting it for now).  The second step, depending upon the game's edition, is to choose their race, and the third is to choose their class.  For a number of people, this is really the same step - the benefits and disadvantages for the different races are often self-selected to match the strengths and weaknesses of the chosen class, so that the character can more strongly excel in their areas of specialization (making sure that Fighters have high Strength and Constitution scores or Wizards have a high Intelligence score, for example).

Choosing the class is, from a mechanical standpoint, the most impactful decision made during character creation, to the point where it is the defining descriptor for that character: "Oh, I'm playing an assassin this time."

EDIT: OD&D had the fighter, mage, and cleric.  Thieves were added in the Greyhawk expansion somewhat later.

In OD&D, there were three classes (I'm using the more modern terminology): fighter, mage, and thief.  They are intentionally generic and comprise a huge variety of stereotypes and associations.  The cleric was the fourth addition to the mix, and it brought with it a huge shift in perception - the cleric's signature abilities, its spells, are drawn from the list of miracles associated with Christian theology - curative powers, blessing the faithful, providing food and water from nothing, raising the dead, etc.

Now, the cleric is billed as a holy leader for any of the different faiths one might encounter within a fantasy world, but a cleric's powers are quite different than one might expect from a Shinto priest or medicine woman.  Using the class to model such a person is whitewashing the game world.  This begs the question: should there be a Shinto priest and medicine person class, then?  If the existing class doesn't work, and we have people with these compelling and very different abilities, shouldn't they have a class?

In other words, should the setting dictate the classes that are available?

The Ferocia have a tradition of spies and insurgents that operate as part of their larger military force, the Custos.  Their skillset is quite similar to that of a thief, with an emphasis on infiltration, assassination, and forgery.  Should there be a Custos class and should I restrict all Ferocia from being thieves, in favor of this more specialized class?  Should I allow other races to become Custos?  Should this class be available for players to use?

The Sohei are a loosely-organized group of wandering warrior-monks dedicated to spreading peace by beating violence out of would-be miscreants.  Through their devotion to peace, they are able to suppress violence in others, but they can fall back upon their considerable martial training if this ability fails them.  Should there be a Sohei class, as Sohei possess a very distinct skillset combining aspects of the fighter and medium (the class of peace-bringers).

I don't know if there is a good answer to these questions.  I know how I used to answer it, but I am not sure how much I still like it.

I believe that a setting must influence the classes.  In the three scenarios I put together for my Sunday players to help determine in which world we will play, one of the biggest ways I introduced the setting to them was by explaining the different classes and their abilities - there was always a fighter and thief-type class, but the supernatural horror setting had both holy figures (luminaries) and diabolic ones (witches and warlocks), and the Shardpunk setting had believers in the old faiths (Adherents).

I also don't believe that a setting should unduly limit choice - I don't support creating a Custos class because that limits the options of Ferocia characters - they can be thieves who are part of the Custos organization, but I do support the idea of a Sohei class as its abilities are unique from any other class available.

My philosophy up until this point has been that if a class would have a unique set of abilities, things that would make modelling it within the existing class framework challenging or impossible, then it should be its own class.  I also, by default, decided that if a class existed then players who had experience with the world could choose it when they made a new character (brand-new players to my world can choose to play either a fighter or thief).

The problem I find is that I now have 14 different classes, most of which cover a very specific, but culturally-important niche within one of the societies in my world.  I contrast this with the fighter and thief, which are these broad, expansive constructs.  If I look ahead, to 2nd edition, I see the cleric, druid, fighter, paladin, ranger, mage, illusionist, thief, assassin, monk and bard (drawing from Alexis again).  Looking further ahead, to 3rd and 3.5, I find the base classes include the bard, barbarian, cleric, druid, fighter, monk, paladin, ranger, rogue, sorcerer, and wizard.  These classes, ignoring the cleric and paladin, are setting-neutral.  They are generic types that can be found in a variety of environments.

I think there are two things at play between how I have designed my classes and how TSR and Wizards of the Coast have designed their classes.  The first is that the standard D&D classes do not provide social status.  Because they are independent of the setting, they gain no cultural or social power within it - being a member of the Tarluskani Military garners significant benefits, but not all fighters who are Tarluskani are members of the Tarluskani Military.  Likewise, being a Ferocia who is a thief (or rogue) does not make one a Custos.

The second, and I think more fundamental difference, is that a class in the standard D&D sense is a game construct - a grouping of abilities that 'make sense' together.  Bards, as supremely skilled artists (chiefly musicians), have a bevvy of music-related abilities.  Rangers, as skilled survivalists, are well-suited to interfacing with the natural world.  My classes, however, have arisen as diagetic constructs - a grouping of abilities based upon the needs of an in-game culture or group.

The challenging thing about this difference is that there isn't an objective answer to this: diagetic classes promote player immersion with the world - a class rooted within an in-game tradition automatically binds the character (and therefore the player) to the world through that institution.  However, a diagetic class lacks the freedom of expression that a class built around a central theme, independent of the world, can provide players - as a Ferocia thief, I can be a member of the Custos, reject them and their methodology, or take no stance for or against them, whereas if I am a Custos, defined by that identity, then I perceive but one of those three choices (as a DM, I would happily support any of the three even for a class named after the in-game group, but I've seen players with similar nontraditional ideas self-steer to a similar class instead).  Names are powerful determinants.

I think an answer might be found in something Pathfinder did, one of their changes to 3.5 that I embraced wholeheartedly - the idea of class archetypes.  Essentially, Paizo defined a group of class ideas (they use the 3.5 list I provide above as their foundation).  Each class has a default ability set but can choose alternate ones providing a specialization around one facet of their class.  Some archetypes were restricted to members of specific races and cultures while others to members of specific organizations.

I want to encourage player choice when it comes to their characters without compromising the immersion that a diagetic class offers.  So having classes represent as expansive categories as possible with culture-specific and group-specific immersive options seems like an excellent way to take the best of both.

I'll have to do some more research and see how I might implement this - at first blush, LotFP resists this kind of complication of base classes, but I am confident that there is a good solution here.

Now, there is one last question that I did not address.  Should all classes within the world be available to players?

No.  I don't think so.  Certainly not at first.  I very strongly believe that beginning players should only be allowed to play classes that are very broad in scope - fighters and thieves.  Once they establish a positive relationship with a member of the class of interest (or a group with a significant population of people with this class), then I'd allow new characters to take on these classes.