It has come up way back when the first 16:10 and 16:9 screens became popular. :wink:

If you want to avoid stretching, you have more or less only these options:

1. Display more of the world horizontally on higher aspect ratio screens.
2. Display less of the world vertically on lower aspect ratio screens.
3. Letterboxing / Pillarboxing.

Or some combination of these. And letterboxing doesn't necessarily black bars, you could just as well blur out the world or have some mysterious fog or something.

You could also define the size of the viewport in terms of its diagonal, instead of either its width or height. Then you are guaranteed to only show parts of the world that lie within a circle of that size.

The most extreme problems arise when you try to display 4:3 content on a 21:9 screen or vice versa, so I would recommend developing for 16:9 and then making it work on other aspect ratios from there.

There are still some 4:3 screens kicking around I'm sure, but its been nearly impossible to buy one off-the-shelf for ages, except for a handful of PC monitors marketed mainly to corporate environments. Back in the day, the question was "how do I design for 4:3 and 16:9?" and now its "How do I design for 16:9 and 21:9?". I think the difference is that 16:9 was clearly going to be where things moved to, but it doesn't seem so clear for 21:9. Oh, unless you're also dealing with iPads, or the myriad other devices as well (though, then you get a pass on 21:9, so far)

But the approach is the same -- either you design such that the extra information isn't beneficial, or you hide it. When you take the design approach, what you're really doing is letter-boxing the design process itself -- you can't place essential information outside the safe area, and you have to think about whether even revealing information sooner to players with different aspect ratios gives them an advantage. Letter-boxing the player's screen is essentially a design-choice to simply ignore these same questions.

Now, something to keep in mind is that this matters most for competitive scenarios -- that's when one player gets a direct advantage over another and fairness is the primary objective. In single-player, non-competitive scenarios, the questions can still matter, but the effect is different; there, your objective is more concerned with ensuring that the pace of progress and level of exploration is not affected. Even in single-player games, though, there can be competitive elements like leaderboards, speed-runs, etc. What's important will depend on what kind of game you're making.

I tend to think, myself, that the best way to approach this in this day and age is to employ a dynamic camera that adapts to a design-safe area that's independent of any particular aspect ratio, and which itself may not match any popular aspect ratio one finds in the real world. That approach doesn't really work for pixel-perfect, sprite-based 2D games, but is a very good approach anywhere that the pixels aren't crucial to the design or aesthetics.

You also have to address on-screen UI in a flexible way -- not only might it change how things are obscured as aspect ratio shifts, but you might also be inclined to deliver different UI properties to tablet users, PC users, and console users anyways. Most serious developers should already have moved to that way of thinking if its important to where they deliver their games.

There certainly are design challenges -- first wide, and now extremely-wide viewports aren't well-suited to games that benefit from a more equitable share of screen area between the horizontal and vertical axis. Imagine remaking Super Mario Brothers today for 21:9 screens -- the change in information presented has a huge impact on design, it forces a decision between revealing probably too much in the horizontal axis, or too little in the vertical axis. Or imagine Sonic the Hedgehog, with its faster, twitchier gamplay -- its even less suitable.

16:9 is only 25% wider than 4:3, but 21:9 is 30% wider again; cumulatively, 21:9 is nearly twice as wide as 4:3 -- the shape of the display area is dramatically different. The side-view scroller has to change a lot to fit in. For other kinds of sprite-based games, say 3/4ths overhead views as in classic RPGs or side-scrolling 2.5D games, they can be made to fit more naturally. I could imagine an RPG that used squat tiles (say 32x24) instead of the traditional, square tiles (say, 32x32) to help compensate.

You can also compensate by adding additional, non-essential-but-useful UI in to letterbox the play area, without adding black bars. Maybe party status in an RPG, or other genre-appropriate information.