salv.bsp (Salvation) by Charlie Wiederhold probably is "the" Quake map with traps. I love the map, many hate it. You should save after each trap/room. I just played it again and died about 10 times. But it is fun. The level also is a "puzzle" level, especially at the beginning you are totally stuck until you get an idea and can proceed.

salv.bsp is in wieder.zip
http://www.quaddicted.com/wieder (neg!ke wrote that description I think)
direct link: http://www.quaddicted.com/filebase/wieder.zip (something 3MB-ish)

Traps can be good but the player should not be surprised by them in a bad way. For example dying without a clue what would happen. A good trap is visible and avoidable with decent playing skill. 

You don't actually need MP1: SoA to play salv.bsp. The only thing from that missionpack is the hammer you get just at the end = useless. I always play it in unmodded Quake. 

forces the player to think and react to the environment more intelligently.

For me, this is the key.

A good trap, or ambush or any kind of gameplay scenario, should present the player with options and consequences. I wrote about this way back in the How Hard Is Too Hard thread.
The ideal you want to achieve is to put the player in a state where they have to 'think on their feet'. This means giving them something to actually think about, time to think about it and time to act. You then have to predict how the player may or may not respond and incorporate those responses into the design, putting appropriate limitations on all of the above. Some of those limitations are to prevent the player responding in ways that render the whole trap pointless. Some may also be to the player's benefit by removing superfluous, possibly confusing information.

Taking your e4m3 example: in that trap, the floor moves slowly enough that the player has time to comprehend the hazard, but not enough time to stop and ponder. One of the things they will understand is that in a very short time the floor will have completely retracted, dumping them inescapably in the lava. But they also understand they have that short timelimit in which to evade - activation of the trap is like lighting a fuse or pressing a stopwatch. Then there are two very strong, though not equally, clear options in response: rush forward to apparent safety, or backtrack to known safety.
Finally, the consequences of that trap are limited to only one negative outcome, but it's big enough ( the biggest in fact, i.e. death ) that no other consequences are needed.

All of these elements are made clear to the player by greatly limiting what else is around. The corridor does not have any footholds, so even before the player has activated the moving floor, the option to evade by jumping to a narrow ledge is cleanly omitted from their options. Lava is a major, well established element of the gameplay, so the consequences of inactivity are made extremely clear in fractions of a second.
The corridor is also very well lit, which removes the stress of uncertainty from the player: everything they need to know is immediately visible so they don't waste time peering into shadows for an escape.

An alternative in this trap might have been to replace the lava with slime and build small steps leading out of it at the far end. This would have made the consequences less negative, but also added more information that the player would have to learn and take advantage of within the timelimit of the trap. Whether or not these differences would have made the trap more or less fun is the sort of decision you as the mapper have to make.

Summary

Good traps have:
a timelimit, time in which to assess and time in which to respond
limited options as to how to respond to the trap
limited consequences if those responses are not used
all of the above communicated to the player in ways that are consistent with what has been established in the game already

Bad traps have:
no way to know what the hazard is before the consequences are experienced
consequences that are disproportionate to what is required to avoid them
confusing information

It's almost certain that exceptions can be cited, but as with all rules of design: you have to understand them before you can break them beneficially. 

other examples of great traps?

In e1m3, the descending roof trap where the NG is. This is very easily avoided of course, because there is only one response and it is plastered in front of the player's eyes even before the trap is sprung. The reason for this is that it is what Valve might, in their commentary nodes, call a tutorial combat. That is, the purpose is not to put the player at genuine risk of harm, but to train them to understand that this sort of trap may be encountered later in the game, and the sort of responses ( shoot a flashing switch ) that they can expect to find.
It is an example of what I refer to above as "establish in the game already."

This is played off towards the end of the level with what is actually a feint. The second descending roof is not actually a trap, and the player is in no danger. But immediately upon the trap being sprung, they don't know that. For a few frantic moments, the player can be expected to look around desperately for yet another shootable switch.

That this actually doesn't constitute a trap does not make it inconsistent with what I have outlined; the second descending roof springs when the player is locked inescapably from the bottom of the shaft ( safely limiting possible responses ), there is very little to confuse the player as to what is happening ( the bottom of the shaft lacks any monsters, impenetrable shadows, jutting architecture etc. ) and what is visible actually clues the player into the gag - the slots into which the two roof halves will ultimately part are clearly visible long before it descends to head height.
And the type of trap is, as I say, established earlier for the player. 

Trapped secrets (in whatever game). 

good thread, and good comments so far. 

interesting, mainly because it breaks all the rules. If some of them were toned down slightly then it would've been much better - i used a rocket jump to cross the lava near the end and get the feeling i missed something there. also the crushers moved a it too quickly to be genuinely fun. but a neat little map overall. 

had some neat "traps". I use the term traps loosely because generally when you flip the gravity of a room the corridor you entered the room from becomes a yawning chasm of death! its a shame the dev team didn't explore this a little more, but it did feel very organic and not at all like "Oh this is a trap segment of the level that I must overcome in order to proceed".

I feel I must mention Ravenholm in HL2 for traps too, mainly because you can turn them against the enemy with very satisfying results, again this feels very organic as you have a choice of avoiding the trap or using it to your advantage.

I think generally speaking the gaming "traps" are evolving into a far more complex mechanic.

1) Early platform games (prince of persia etc), Trap = insta death.

2) Doom/Quake/Duke etc. Trap = avoidable but usally high damage or death.

3) half-life/deus ex/etc. Avoidable traps that can be manipulated by the player for better or worse.

4) Current generation, Traps can be avoided/moved/used against attackers/ reconfigured in some way. *PHYSICS*

5) Next gen, who knows :) 

Falling into a "trap" will actually increase your health and kill your enemies, because all games will cater to the lowest common denominator, who get all confused and angry when a game "beats" them.




/only joking

/(well nearly joking) 

Prince of Persia was actually ahead of its time, becuase it was the game I remember where the bad guys obeyed the same rules that the good guys did. And you could definitely use traps against bad guys -- the spike pits, and metal guillotine things would both kill a bad guy if you could force him into it using your sword skills (when they blocked your swing, it pushed them back a little bit.)

Actually Prince of Persia and Wolf3D and Doom all impressed me with this. Bad guys got killed by the same stuff as good guys, including traps, friendly fire, etc. They also left dead bodies in place rather than fading them out like other games I'd played. I wouldn't call this "realism" so much as a more interally consistent game world. 

becuase it was the game I remember...

becuase it was the first game I remember... 

i really like traps that can kill enemies - in the early original levels, maybe ep1 or 2 there's a big spike that you can gib an ogre with if you're canny - great fun. I've got a few that hurt enemies; blastertraps that fire the length of the corridor, but its almost impossible to kill em with this. I've also got quite a few sections where you can lure fiends + spawns into jumping into the void, not too complex but i've ended up putting more fiends in the outside parts just to take advantage of this.

i also remember the bit in ravenholm where you climb on top of one of the car / crusher traps when it rises in order to progress - a bit like zelda the ocarina of time in the shadow temple.

one idea i had was to create land mines that the enemy could trigger if they blundered over the top of em or the player could trigger by shooting. Would be cool but pretty complex to set up, probably with a number of switches nested inside each other plus a few relays.

informing the player of how you world works through traps is a good idea too - an example of this with the landmines might be having racks of inactive ones at the start of the level, followed by the first active one being triggered by a monster - but not injuring the player, the ones after that being the 'real' traps. i'm working in qouth so maybe its here that i can use the info_bomb (not sure on the name) or else the notnull explosion hack.

metslime - because ;)