What follows is descriptive, not prescriptive. I am attempting to describe how I am designing and running a dungeon, not how you should do it. Most weeks, I am scrambling to get ready for game night in between family, work, and my music obligations. So with those caveats, here is my approach:
Campaign Mindset: From the start, my mindset has been to run it as a campaign setting and not an adventure module on steroids (which is how most published megadungeons appear to me). Rather than focus on rooms, I focus on areas, peoples, etc and then fill in the details as needed (or as I have time), just as if I were running a typical non-dungeon campaign.
My Dungeon: The Montporte Megadungeon is designed by me for me to run. And I run it with people I know. I am not creating something for publication or something to even post on my blog for others. It is very specific to me and our Monday Night Gaming Group. I don't feel like I have an audience of gamers/consumers to please, just a group of friends to entertain.
As Needed: I only create the dungeon a little in advance of where the players might go. At the beginning, this meant a fair amount of mapping. But now, not so much. There are several reasons I do this: (1) I don't have the time or energy to create a 20 level dungeon all at once; (2) I don't always know what will catch the players' interest and I would rather have things be more open-ended and flexible; and (3) Inspiration comes to me slowly and if I try to do too much at once, the dungeon turns out bland. I need to give myself time to let ideas germinate.
Mapping Part 1: Fortunately, I love to draw dungeon maps. Unfortunately, we are playing online with a VTT (our group uses either Roll20 or Fantasy Grounds, depending on the GM...I opted to use Roll20). This means that I could either draw maps by hand, scan them, and upload them...or create maps digitally. I opted to create them digital in AutoREALM, which is a slower process for me, but it cuts out the conversion process.
Mapping Part 2: I usually start my mapping process of a level by creating a 5x5 grid in Publisher. Each square in the grid is 10", which means I use the custom setting to create a 52" x 52" page (allowing for 1" margins). I then place a text box in each grid, describing in as little detail as I can, what is in each square. In dungeon terms, each square represents 300" x 300", the same sized used in Stonehell. I then, as I need to, create the detailed 300' x 300' maps in AutoREALM, with a 5'/square scale. I convert it a JPEG and--Presto!--I have my map for Roll20. I can also drop the JPEG file into my Publisher grid and create level map, which can also easily be converted to JPEGs. By using blackened rectangles to cover unexplored areas, I can create overview maps for the players like the one below (it takes about 15 minutes):
|Level 1, From the Players' Perspective|
Tropes Part 1 ("the lack thereof"): For the most part, I have avoided the OSR cliches and tropes--random tables, lots of slash and grab sessions, and traps for the purposes of having traps. It is not that I have a philosophical or ideological reason to avoid them. My reasoning is more pragmatic: It takes a surprising amount of time to create a random encounter table that might only be used 1 or 2 times. It is easier to just create the encounter than to create the table to generate it.
Tropes Part 2 (my trope): My one conceptual trope has been the computer game, Myst. I am not trying to recreate Myst as we play, but I like the idea of using interlocking clues and details so that the players gradually piece together the multiple back stories of the dungeon. There are a few chatty NPCs in the dungeon, plus lots of scrolls and documents. In fact, every encounter is a clue to the dungeon. Not all clues are earth shattering, but they all pile on one another to paint a picture and tell the story of the dungeon.
Improvisation and Planning: I only occasionally use set-piece encounters that are tied to specific physical locations within the dungeon. They are there, but there are not a lot of them. I do, however, improvise a lot. For a lot of people, improvisation implies a lack of planning or forethought, kind of like "winging it." As a musician, improvisation means something different to me. Improvisation means creating responding to others as you play, but it still requires as much planning, forethought, etc. I spend a considerable about of time planning for each session, but I look at options the players have in front of them and how they might respond, rather than planning specific encounters ahead of time. I am not sure if this makes sense to you, the reader, but I have found it to be my style of GMing...at least most of the time.
Rationalization and Logic: The Montporte Dungeon has a series of backstories that provide an inner logic for myself and the players. However, it only works when we (our group) let go of (1) the question of why a big multi-layer hole in the ground would be there in the first place; and (2) previous preconceptions about megadungeon play (this has been a bit of a struggle as the cliches of dungeon play haunted us when we started).
Inspiration: I read a lot--close to a book a week--and while I do not plan my reading around gaming, I often read things that end up in the dungeon. A recent example are the small bottles of impossibly cold water (why did it not turn to ice?) found in Session 22. These were inspired by So Cold the River, a novel by Michael Koryta. My days with the Appendix N types of literature are mostly in the past, but I am finding inspiration in unlikely sources. This has proven to be a benefit as some of the guys in our gaming group have way more gaming stuff than me. I have thus far avoided the dreaded "You pulled this from B1, didn't you?" or "Hey, this sounds like Dragon #135."
Further Reading: John Arendt (Dreams in the Lich House) recently published two posts on megadungeons that present an interesting point and counterpoint: You Will Never Finish That Dungeon and From an Alternate Universe. I found them interesting and helpful. For me, Michael Curtis' (The Society of Torch, Pole, and Rope) post, With New Old Eyes, is my "go to" reading for running a megadungeon. Peter D (Dungeon Fantastic) wrote an excellent post, Megadungeon Play Reflections-The Immediate and the Cumulative, that lives up to "Fantastic" moniker. I have previously tried to capture a bit of my own thinking on designing and running megadungeons: Five Paths For Dungeon Design and Creating Meaningful Choices in a Dungeon-Centered Campaign.