Any good SF world must be understood as a possibility space. A universe which is characterized first and foremost by its physics:
By what is and isn't possible, and by how easy (or difficult) each of these non-impossible things are to do.
Once you have created the possibility space, you must extrapolate how the people in the world will exploit the possibilities. Because, you know, they're people. They are going to exploit it. Some of them, at least. That's what people are like.
I'll mostly be talking about science fiction in this post, but the principles I refer to are eminently applicable to fantasy as well.
For instance, can spells of healing be cast, in this world? If yes, how easy (or hard) is it to learn to do so? Once you've learned them, how easy (or hard) are they to cast? How much can they heal? Which factors, metaphysical or social or other, act as a "brake" on their usage? Which factors acts as a "brake" on the dispersion, over time, of the ability to cast these spells? What about other types of spells? Are most spellcasters generalists or specialists?
Fantasy has a well-deserved bad reputation, for using magic as a plot device (instead of an integral part of the world metaphysics). Magic can always do exactly what the author wants it to do, and only when he wants it to be able to do it, and nothing else. It's a shitty and retarded attitude, one that needs to be resolved.
There's a wonderful example, in Ursula le Guin's "Wizard of Earthsea", of what a world in which magic is possible, is a part of the possibility space of the world rather than a convenient plot device for an intellectually challenged author (or GM), would be like: The scene with the spell-controlled rain cloud.
Again: Can faster than light travel be done, in this world? If yes, what's the process like? How expensive is it? How dangerous is it? From where can you travel, and to where can you go? Nevermind the made-up physics of how. Focus instead on FTL being a possibility within the world, and on the circumstances of FTL.
Audiovisual science fiction, movies and TV shows, also has a bad reputation, and equally well-deserved, but in its written form, which is very much its true form (science fiction is something you read, and by that I don't mean media-tie in material such as Warhamster 4*104 or Star Drek), things are often better.
Star Drek, by the way, is an excellent example of the script writers failing to understand the show's world how it ought to be seen:
As a possibility space. They keep introducing wonderful technological devices to conveniently resolve plot situations, and then completey forget about them again, even thought according to the objective and universal principles of psychological realism, such technology will be used and re-used, again and again, as the people in the world (at least some of them) exploit the possibility. I often use the phrase treknology to mock this particular category of incompetence.
About a decade and a half ago, I came across the 2nd edition of the GURPS Space RPG supplement. Perhaps surprising many, GURPS Space1 is not a world book, a supplement that defines a setting for campaign use, but instead a worldbuilding guide, to help the GM build his own setting.
It was very helpful, opening my mind to a wide variety of possibilities, by asking me interesting questions. For instance, it points to the distinction between moving matter at FTL vs moving information at FTL. What if the world does have FTL travel but not FTL radio? That means you'll have to use courier ships to send messages, or at least robotic probes.
And that affects how the world will be.
Accepting that as a world fact, you extrapolate from it, and figure out how the resulting societal organizations can be. Because absent FTL radio rules out a great many things. Some well-known examples of FTL courier worlds are the Traveller RPG universe (available for a variety of RPG systems - offhand I'd recommend GURPS Traveller or Mongoose Traveller, as being the least bad of the available ready-to-use options), and the universe in Timothy Zahn's Conqueror's Trilogy.
(A reverse example appears to be the world of Orson Scott Card's "Ender" tetralogy. FTL radio is clearly possible, via a so-called ansible, a term Card borrowed from Ursula le Guin's works, but FTL travel seems to take a lot of time, although thanks to Einsteinian time dilation (possibly exaggarated by Card - as I recall it, he's not clear about the details) the travellers themselves experience only a few weeks or months of time, even when travelling very long distances.)
So, FTL radio or not. The difference matters. The world builder, whether a GM or an author, should choose the option that leads to the kind of world that he wants. If the captain gets in trouble, can he phone home?
Same with FTL travel. What if it's only possible via huge stargate-type structures, that you must first build and then move, at STL speeds, to the destination systems? As in The Algebraist by Iain M. Banks? EVE Online, the famous MMO, also has stargates, but it's not clear from in-game how they're constructed, and I frankly can't be arsed to read the many "lore" articles and short stories that are available. EVE isn't an RPG anyway.
Certain kinds of worlds emerge from such a physics. Many other kinds of worlds can't emerge from such a physics. It's very possible to objectively fail, by depicting a world that is contrary to the physics postulated in the story.
What if FTL travel can't be initiated close to large celestial bodies? That means you'll have to travel, at STL (using some kind of reaction drive) away from the planet or moon you've launced from, for some hours or months, before you can engage.
What if FTL travel can't be ceased except close to large celestial bodies? That means that you'll have to "come out" of hyperspace near a planet or sun, or maybe a moon, but if you want to visit a faraway asteroid, you're going to have to slowboat all the way. If you can't change direction after you've initiated FTL, then you'll have to aim very precisely indeed2, otherwise you'll miss the target sun, and instead of dropping out of hyperspace like you wanted to, you'll keep going. And going, and going.
The strategic implications are interesting. If you have a solar system state, or an interstellar state, how do you guard it against possible invasions? Where can visitors, hostile or not, arrive? If they can arrive anywhere, you'll have to guard valuable attack points, such as planets, well-colonized moons, important space stations. But if they can only arrive at certain points, no more than 5 or 30 or so per system, then instead of guarding your valuables you can guard your gates.
Another possibility is warp points orbiting the stars. Perhaps with each solar system having about a dozen of these. The limtiation could be that you can only enter hyperspace from those, or can only exit hyperspace from those, or both. Or there could be separate entry warp points and exit warp points.
Again, you have something to guard, strategically, but it'll be interesting how easy or hard they are to find, and from what range you can detect them. For instance, can you detect such warp points from a distance of a few light years? A couple dozen light years? Or only at a distance of a few AU? Can you detect them from inside hyperspace? Are such warp points permanent features, persisting for billions of years? Or do they change? If so, do they change on a scale of millions of years, thousands of years, decades, or hours? Do they change by appearing and disappearing, or by changing properties? How predictable are such changes?
What if most "standard" solar systems, with suns similar to ours (yellow dwarves), have not about a dozen such warp points, but most likely only one or two, or sometimes none? And if they can collapse randomly, and not re-appear for thousands of years?
Consider how that affects interstellar exploration. You can get to Alpha Centauri, by travelling from one of Sol's warp points (which, inconveniently, is some dozen AU out, currently close to Neptune, so you'll have to take a couple of years getting there, even using an ion drive), to the one warp point there that has been detected using long-range scan, but what if, after you arrive, the warp point disappears? The only warp point Alpha Centauri has. Then you'll be stuck, forever, just you and the other 2-4 dozen crew members of the exploration ship, and your 10 years of supplies. Are you brave enough to take that risk? What if none of the Western nations can find any volunteers, NASA and ESA coming up dry, but the Chinese can? Even if they're the kind of "volunteers" that inarguably belong inside scare quotes.
Or what if a standard solar system has not about a dozen such warp points, but about a hundred, or several hundreds? Or if the warp points don't orbit suns, but rather drift through space according to another principle, so that from the frame of reference of any one solar system, they appear to be passing through the system? Perhaps at thousands of kilometers per second?
Such things matter.
Fuel for FTL
Does FTL require some kind of fuel? The physics of the Traveller universe are such that enormous quanties of pure hydrogen are required to make jumps. Nevermind why that is so. Focus on the consequences, the world implications of it. Starships must devote a significant percentage of their volume or mass (depending on which RPG system spaceship design subsystem you use) to FTL fuel tankage. The longest you can go in one jump is 6 parsecs (19.5 light years), as an absolute and hard limit, and a starship with such a long jump range is an extremely specialized vehicle. Each "jump", whether at the starship's maximum range or shorter, takes one week, plus minus maybe a day and a half (the GM rolls randomly).
Military ships in the Traveller universe, cruisers and so forth, usually have a more modest range of 4 parsec (compared to usually 2 parsecs for trader ships and scout ships), which still requires a lot of the ship's capacity, leaving little for STL engines, weapon systems and so forth. This has the consequence that equivalent-sized (and -priced) non-FTL capable ships, such as the famous System Defense Boats, are much more powerful militarily. Faster, better armoured, and with stronger weapons. A 400 "ton" system Defense Boat will routinely pwn a 500 "ton" starwarship, and maybe even a 600 "ton" model too. (Another example of this principle are the Hellburners from Cherryh's Alliance/Union universe.)
Refuelling is fairly easy, in Traveller. You just dive through the upper atmosphere of a gas giant, scoop up lots of gas, fly out again, then your onboard refinery refines out the hydrogen in some hours, or some weeks, or however long it takes. Then you can jump again.
Or you land on an Earth-type planet, and use electrolysis to extract the hydrogen from sea, lake or river water, or from a glacier or polar ice cap. Or you can slowboat it over to the nearest comet, and again do the electrolysis thing, or perhaps get hydrogen from it in other ways (such as from frozen methane).
In Zahn's Conqueror's universe, FTL travel can be done from any point to any other point, but it requires expensive fuel. Far from as bulky as in Traveller, but costly to produce. Furthermore there are only two possible FTL speeds, one standard speed at about 3 LY/hour that is used by most starships, and a faster speed twice that, only achievable by small ships, space fighters, scout ships, and courier ships. The fuel needed for this faster kind of FTL travel is also stated as being five times as expensive as the regular FTL fuel, so (as depicted in the first book in the trilogy) it's a non-trivial decision for a small colony world whether to send a courier ship to the nearest more populated planet, to report certain anomalous events.
Whether or not the "tachyon wake trails" can be detected, at a distance, and how easily, and how precisely the direction vectors of such trails can be measured, and how they can be obscured, is also of great strategic relevance in this world.
Of course, any such universe is going to have mail service. Traveller has its X-Boats. Zahn's world no doubt has regular courier ships, or something like that. Messages will be sent. The problem is only with urgent messages, and that tiny remote colonies and outposts won't get visited by the postman very often.
Above, I talked about landing on a planet, in order to refuel via electrolysis. But of course that's often non-trivial. Gravity is a total bitch. Once you've made it from the surface into orbit, you usually don't want to take your entire starship down there again. Ever. Even with advanced high-thrust high
Many large starships use auxilliary craft to move crew and cargo down from orbit to the surface, then back up again. Some writers and GMs do that because they see it as tradition, how it "ought to be done", rather than because they have even a minimal understand the underlying physics.
STL drive systems basically fall along a spectrum, where at one end you have high delta-v, often hundreds of kilometers per second, or even tens of thousands (and yes, that is a bit fast, relative to c), but very low thrust, perhaps only a fraction of a milli-G. The ion drive concept, mentioned earlier, is one such example. Also some antimatter based drive systems can be like this.
At the other end of this spectrum, you have drive systems with low delta-v, just a few kilometers per second, but with high thrust, often well more than 1 G, meaning you can thrust directly into orbit, rocket style, if you have the required delta-v (11.2 km/sec, for Earth). Another option may be to fly from surface to orbit, using aerodynamics (wings, or a lifting body), but you're not going to do that with the puny mG of an ion drive. These kinds of high-thrust drive systems can be chemical rocket, or fission or fusion rockets (or pulse drives), or antimatter rockets. As seen by the low delta-v, they use a lot of reaction mass.
Each extreme of this spectrum is optimal for different things, and either crappy or downright useless for others. High delta-v systems are good for long-distance travel and not needing to refuel often, e.g. huge spaceships, mothership-style ones. While if you want to move stuff from surface to orbit, or the other way, you want high-thrust systems. So the obvious solution is that the huge starship carries a few high-thrust shuttles, as well as plenty of reaction mass for them.
(Military combat ships want high-thrust, usually, e.g. if I were to design a Traveller-style System Defense Boat then I'd give it a high-thrust drive system, while I'd give a low-thrust system to the much larger starship designed to carry the SDB. And similar, if you can find a way to push space fighters into your science fiction world, you're going to want a high-thrust drive system for them.)
Atmospheric flight in itself is also tricky. To have a vehicle that can do that, you need all sorts of control surfaces, wings with flaps and so forth, tail fins, you need to worry a lot about aerodynamics, and there may be thermal effects of the air flowing over the surfaces at very high speed. So you jump through those hoops when you draft your shuttle, making a bunch of design compromises to get that capability, which you can skip it for the huge starship.
The X-boats from the Traveller univesse, the mailboats, skip STL entirely. They have no STL drive systems of their own, but only FTL engines, in order to make the postal service as cheap as possible. Again, it's a case of the in-world designers seeing a way to save money and resources, by gimping the ship they're designing in a way that won't matter to its intended function.
Thinking about all this leads to a better world. It leads to specialized spaceships. A world in which all spaceships can routinely do everything well is a boring world. There's much more dramatic potential in a world in which strict physics force spaceship builders to make hard choices, to weigh options, to build specialized ship, ships optimized for various things.
Of course, at extremely high "Tech Levels", such as depicted in Iain. M Banks' "Culture" universe, it makes sense that spaceships are just ships, and that they can all do everything equally well. It's the logical conclusion extrapolated from ever increasing scientific understanding and engineering expertise. But it's also boring. It loses dramatic potential, meaning that you'll need - badly - another source of said potential.
Back and forth
Of course, that's not how I do it. Some authors or GMs decide on a physics (or metaphysics) premise, then extrapolate from it, and that's that. That's fine. It's admirable, deserving of much respect. And sometimes it even leads to worlds that are in themselves interesting, rich in dramatic potential.
But what I do, is I start with a physics (or metaphysics) premise, and extrapolate from it, examining it from the in-world point of view, a world populated by Humans (or other people), who are by nature horny, afraid and eager to achieve material wealth. The physics is a set of possibilities, and those possibilities will be exploited.
I then look at the world that emerges from the physics, and I evaluate it. One of the criteria I use is, obviously, dramatic potential. Usually what happens next is that I find that I don't actually like the emergent world much. Sometimes I discard the initial physics completely, as being an inherently bad idea. More often, though, I go back and make a few tweaks to the physics. Then I extrapolate from these new physics. See if the new emergent world is better. Often it is. If it's good enough, then fine, I have something I can use. If not, then I go back and apply further tweaks, until I either get something that is good, or until I realize that the general concept behind the physics I have in mind is flawed, and that no good world can emerge from it.
Peter Knutsen typed these letters
1. At least from 2nd to 4th edition, the ones I've read, but presumably 1st edition too.
2. One light year is about 9.5 million million kilometers.
2. One light year is about 9.5 million million kilometers.