Well like I said, this is my personal taste. It is by no means an absolute rule. And I agree that some empty rooms can serve a purpose. But more commonly, they tend to dominate bad maps, where empty rooms and corridors are all over the map. I believe that a map that is more densely packed with things of interest, make for a better dungeon crawling experience. This doesn't mean that every room must be stuffed with monsters, items, traps and puzzles. But I think it makes for a better experience if there's something to see and do in nearly every space of the dungeon, even if it's something trivial.
Depends on the dungeon and-or the PCs' reasons for being there, too. As a starter adventure I've had parties go into (in theory) cleared-out dungeons with their goal being simply to map the place, as a training exercise. Sure there's some minor monsters have moved in, but on the whole the place is mostly empty. (and yes, this is an unusual type of adventure)
If a dungeon has multiple bands of warring occupants, empty rooms could also represent a buffer between their territories.
Of course a dungeon can have modifications that weren't part of the original fictional design. When I talk about designing with a purpose however, I mean the purpose that the DM has in mind for an area (and not the fictional person who originally created the dungeon). A collapsed corridor or tunnel of a burrowing creature is a purposeful area. There is a reason it is there. But more often than not you'll see a lot of rooms and corridors that don't seem to serve any narrative purpose at all. This doesn't mean that every room MUST have a purpose. But in my experience, when you think about the purpose of some of the areas, you end up with a better result.
Ah, I think we're coming at this one from opposite directions.
When I'm designing an adventure site my first thought is "what was its original in-fiction purpose" (or "why is it here"), followed by "what has happened to it since, if anything". Only after that will I overlay any purpose I-as-DM might have for that area now, as in "this was once the main hall, it's the biggest room, so that's where the Ogres are gonna live".
For example, if you're designing the lay out of a castle, how far would the kitchen logically be from the food supplies or the dining area? How far would the servant quarters be from the location where they do their job? When you think about these sorts of things, you gain an understanding that these areas should logically be closely connected to one another.
Agreed re the kitchen-dining example. Servants' quarters ofetn weren't all that close to their work site - kitchen staff at a mansion who spend their days in the basement might have their rooms in the attic, for example; or even live in a detached barracks.
This could mean that as a DM you're designing a whole lot of rooms that will never be explored.
Not a problem.
Having 'some' optional rooms is fine. But if the players take a right turn instead of a left, and end up skipping half the dungeon, then there's surely something wrong in the design.
If the players/PCs aren't thorough in their explorations, that's on them. All they need to do is map the flippin' place and they'll soon see the unexplored passages.
I think there are plenty of reasons why a dungeon might have only one way in. Especially if the dungeon is supposed to be some kind of cave or tomb. But I agree that it is a thing that DM's should give more thought. Plus there might be alternative entry points that weren't part of the original fictional design. When a dungeon has more than one entrance/exit it gives the players a strategic option to consider. I would however suggest then that it should be a meaningful choice. The entry/exit points should be notably different, and have their own advantages/disadvantages. As a DM I also like to include an obvious route and a not so obvious route (for example, an underwater passage that allows for a more stealthy approach). A DM should also be cautious not to overdo it. If the players have too many choices, their point of entry may feel irrelevant.
My go-to example for this is the castle/dungeon part of L1 Secret of Bone Hill.
There's about six different ways to get in to that thing, some obvious, some not so. What this allows for is if a party gets beaten back from one entrance they can try another; and-or they can scout to find an entrance that at least appears less risky. (and as DM it's great if you want to run the module more than once over the years, as the choice of entries means no two parties will ever approach it the same way!)
I'm not too concerned about the tactical advantages/disadvantages of each entry point - they are what they are and the players/PCs can approach them as they like. They're not likely to have much information on what's inside unless something's telegraphed or (very rare IME) they're willing to spend the time to observe the place for a few days, thus making it quite intentionally difficult to make a "meaningful" choice* except by trial and error.
* - other than obvious challenges such as climbing if, say, they decide to go in through a roof or high window.
Example: in Bone Hill the obvious means of entry to the castle are through the front door or by scrambling through a big hole (or two?) in the wall around the back. There's telegraphed dangers to the front-door option (obvious circular char marks left by the fireball-happy wizard in the place), but the hole option is also quite risky. There's also upper windows you could go in; if memory serves there's also an entry through the roof; and at least one secret entry through an outbuilding direct to the dungeon below.
Once inside, Bone Hill also has several loops both vertical and horizontal, though there are a couple of dead ends in the dungeon.