Ruined Cities Are Really Hard

Or, dealing with hyper-dense 'dungeons'.

A lot of my stuff revolves around dead cities, ruined cities, cities where something went wrong. This is me putting some thoughts about such environments being really hard to play whilst satisfying some of the stuff I like to do.

See the city. It is dead, filled with buildings bereft of their original purpose, re-imagined as lairs, traps and storehouses for treasure. Compare this to the traditional dungeon - each room has specified exits and entrances, whereas the city offers a practical infinity of entrances, exits, and approaches. This, in addition to the sheer sizes, is a problem to be solved. Two main approaches spring to mind.

Abstraction.
Movement through and the contents of the majority of structures are abstracted, often through the use of procedural generation (this house has *dice dice* nothing) - those structures which do contain items of interest are 'zoomed' into, breaking away from the strategic (travel-based) and moving into the tactical, individual level movement, most obviously combat. This is intuitive, and means the game isn't a slog of this house is empty, after the players describe surrounding yet another ruined manor. However, such zooming immediately informs players that something interesting is about to happen, whether this been combat, traps or a secret to be discovered, meaning they will deploy in a manner to take maximal advantage of the environment. (More on environment usage later.)

The characters, assumably, will be moving and acting in a far less cautious manner during standard travel. (This could, of course, be considered in the abstraction, moving far slower in the strategic view.)  This effect ruins the opportunity for players to be surprised - although, a solution for this would be utilizing more active opponents, who attempt to engage from surprise, forcing the players into positions less advantageous as they are the defenders, adapting to the situation as dictated by the ambushers. Such an addition rewards players defining themselves scouting and planning for such situations, dictating a marching order taking advantage of the nuances of the specific buildings and streets in the encounter area.

This, however, runs into another issue within abstracting the dense urban environment. Using generic floor-plans and streets leads to strings of encounters effectively occurring within the same environment, a street lacking in interesting nuance, with the same layout of buildings offering the same opportunities. Generating an interesting and unique street and/or floorplan(s) however, is going to take time - the opposite of what a surprise engagement offers. Building a large library of interesting nuances yet somewhat universal layouts would negate this somewhat - but then the difference between the abstraction and a complete mapping shrink, reaching the point where complete mapping might make more sense. The balance between a nuanced and interesting engagement locale with the speed of the generation is very hard to strike.

Complete Mapping
Completely mapping a dense, decaying urban environment is a gargantuan amount of work, which immediately makes this option less appealing. Even ignoring this significant limitation, we run into the fact such a huge amount of information is really hard to use at the table. Each structure would require some form of representation, informing (or inspiring) the GM as to the external and internal structure of the building. This could be achieved through some form of short-hand tags or keywords, the combination of these phrases rapidly building a mental image to be imparted to the players. Such a system would require a degree of training in the GM, even just to simply learn this skill. The advantage of such complete mapping is the ability to instantly determine the form and nuance of the locale an engagement is occurring within - the GM will know through a system of tags there is a barricade which offers either side an advantage, without the need for a potentially cumbersome or slow generation system. The key to achieving a working Complete Mapping is a really effective manner of splitting information into table-usable chunks, with both player-facing and GM-facing maps and information available.

Both are hard and leave me wanting somewhat. Whadda you guys do?