(In short, down, when under thrust, is determined by the direction of your drive axis, and specifically, is the direction your engine is in – because it’s pushing you the other way.)
I feel somewhat bad, sometimes, for violating this one here and there, sometimes quite egregiously, but there’s a reason for that, and it has to do with the fiction in my science fiction. Specifically, the inertial dampers that I talked about here. See, way back in the day before those were invented, from the Phoenix stack on up, spacecraft and then starship design was indeed constrained this way; virtually all designs were tail-landers, with the decks perpendicular to the drive axis, and the only exceptions to that rule being belly-landing vehicles intended to operate in atmosphere and/or land planetside, in which case the need to deal with the planetary gravity field took precedence and those aboard pretty much had to suck it up and deal with the inconvenience when in space and under thrust.
(We omit, for the moment, the complexities of spin gravity and the combination of spin and thrust “gravity” that gave the world terms like thrustdown and spindown and realdown and lots of rather complex gimballing mechanisms.)
But then the inertial damper was invented, thus ensuring that the interiors of starships equipped with it were under microgravity all the time, even when they were under thrust, and naval architects almost immediately split three ways:
- The traditionalist school, who had been building tail-landers for a millennium and dammit, were going to keep on building tail-landers, because that’s how spacecraft ought to look, and for that matter, it’s kind of nice to still be able to fire up the drive if your inertial dampers break down, isn’t it?
- The convenientist school, who countered that people had been complaining about what a pain in the ass ladders, companionways and elevators were for getting about inside spacecraft for most of that millennium, especially if you’re not under thrust most of the time, that long corridors are much nicer, and that it’s good not to have large pieces of equipment split between a half-dozen decks, and so now that they could build starships as belly-landers, dammit, they were going to.
- And the spacer school, who pointed out that if there’s one thing that you could learn from modular and beehive habitat design over all that time, it’s that “down” is a strictly local phenomenon and one only useful under a few circumstances anyway, and that in a microgravity environment not only can you arrange your decks any damn way you please, but you don’t even have to be consistent in doing so, and proceeded to arrange their designs’ interiors in whatever way they felt was useful at the time and place.
In what I think of as the modern era, the spacer school has essentially won the argument, although examples of the other two schools do still show up. (After all, planet-landing craft have to be consistent one way or the other, what with that planetary field to contend with.) Among people who have the relevant technology, at least – the constraint still exists, and applies in full to anyone who doesn’t have fancy ontotechnological physics-editing tech to play with.