DCSS: Successful Open-Source Game Development

[rosstin]

Hey all, this is Rosstin. I'm writing a Gamasutra article about DCSS as an example of successful open-source game development.

EDIT: It's up!! If you have an account at Gamasutra, please comment! Hopefully I can get featured.
http://www.gamasutra.com/blogs/RosstinM ... uelike.php

I'm posting here to get some opinions from the dev team about what the successes and hardships of the open-source model have been for Crawl. I'm also curious if anyone who posts here was a member of the team which approached Linley in the 2000s. Galehar seems to be the most active dev poster. I'm curious why Crawl was chosen for further development in particular. What experiences have you had working with a codebase that you didn't originally write?

I think one of the most interesting things about Crawl's development is the philosophies behind it. There's even a philosophy page in the game's help menu. As far as I can see, the core dev philosophies are:
# Don't waste the player's time.
# Optimal play should also be fun.
# Every decision you make should be meaningful.
# A variety of strategies can be used to succeed.
# Elegance in number of features and options

I'd like to hear some dev opinions on whether these are accurate, what you disagree on, etc.

Essentially, if the devs want to vent or reminisce, I'd love to hear it!

Me: I'm a game developer and avid Crawl player. I've ascended a number of times although I've never gone for the 15...

I made this for the article so far:

[rosstin]

What do you think are the largest key improvements that have been made from Linley's Dungeon Crawl to today's DCSS?

(Also, if anyone knows where I can get a version of the original Linley's Dungeon Crawl for comparison purposes, I'd appreciate it. I have this: http://crawlj.sourceforge.jp/down_e.html . Is this essentially the 2004 version?) NM, I found one: http://db.tigsource.com/games/linleys-dungeon-crawl

[galehar]

Galehar seems to be the most active dev poster.

I used to be a quite active poster but not so much these days. I think mumra posts more than me. And I don't code much either, I'm on a break.

  Code:

I'm curious why Crawl was chosen for further development in particular.

I chose DCSS because it felt like "the modern roguelike" at the moment (2008). Now, it feels so old
I went straight from nethack to crawl. Loved the modern interface and the great balance. And the lively team with constant addition/cut/changes.

What experiences have you had working with a codebase that you didn't originally write?

You learn a lot. Especially with crawl which as such a huge code base, and with so many different coding styles. You learn as much about what to do than about what to not do.

I think one of the most interesting things about Crawl's development is the philosophies behind it. There's even a philosophy page in the game's help menu.

Yeah, this is great. Dpeg wrote this and guided the development team for years. But he's much less involved now. He will still probably come around talk to you about it
With the constant turnover in the devteam, the vision is evolving, but I think most of the stuff you listed is still valid and it's keeping a general direction for the game design.

What do you think are the largest key improvements that have been made from Linley's Dungeon Crawl to today's DCSS?

The first major ones were autoexplore and autotravel. This is what created the DCSS fork.
Tiles and webtiles interfaces I'd say. There's even an android port, but it's a bit rough.
All the new gods, and many of the old ones redesigned (thanks dpeg!)
The new skill interface (I might be biased about this one )
There's so many new content since Linley's crawl: several branches, many monsters, species, mutations, hard to decide. Maybe the changes to spells have been the most important ones. The additions and the removals. A bad monster can be annoying, but a bad spell can break the game.

[rosstin]

Thanks Galehar! Your input is really appreciated!

What do you think are the key features that set DCSS apart from other roguelikes?

[Speleothing]

I would say autotravel/autoexplore are extremely important DCSS features. Without them, the game just wouldn't be fun to play for beginners. Crawl is capable of drawing a much broader & more casual audience because of them. I would have never played without them.

[rosstin]

I remember somewhere there is a list of most popular Roguelikes by year. Anyone know where that is?

NM, here: http://roguelikedeveloper.blogspot.com/ ... date=false

[rosstin]

Thanks Spel! This is the kind of thing I need to hear, the fact that you wouldn't have played Crawl without auto-explore is important.

I'm also curious from the devs-- MarvinPA mentioned that having a larger than normal number of devs has proved surprisingly beneficial to Crawl. What was the original team size? I understand that Linley created Linley's Dungeon Crawl, and then 2 devs contacted Linley (and some other developers) to update the license around 2004 (?). I presume that those 2 devs then expanded the team of core devs. Who were those original 2 devs? Are they still part of the team? How many devs are active now?

Also, who wrote the "philosophy" section of the DCSS manual? Was that DPEG?

[reaver]

I love Gamasutra and am glad you saw fit to have a rougelike article.

Since you're looking at why Crawl Stone Soup was successful, you might want to look up Brent Ross's Crawl Stone Soup 4.1 - a similar attempt to continue Crawl development that failed, sparking the Stone Soup project. An account of what happened is here: http://crawl.develz.org/wordpress/the-d ... stone-soup .

According to your article:

"The distortion weapon being a club, I was also unable to chop up corpses for food, and my character began to starve. I had a choice: I could search the dungeon for a spellbook containing Animate Skeleton, which has the side effect of generating chunks of meat from corpses, or I could unwield the weapon. I chose to attempt to remove the weapon, and was unluckily sent to the Abyss, where my character eventually died."

While this is an excellent example of the kind of gameplay Crawl provides, the whole "you can't butcher while wielding a blunt weapon" thing was recently removed from the game. While the 0.11 release had it, 0.12 doesn't. I don't know how important this is for your article.

The definition of a roguelike is something that's been struggled with for years. The Berlin Interpretation gets throw around a lot, and Temple of the Roguelike has a robust definition I personally like.

At the point where you discuss how Nethack was succeeded by Nethack 4, you might want to talk about other forks like UnNethack, which is also actively developed and has more changes than NetHack 4 (like new dungeon branches).

Currently there are 19 official Crawl Stone Soup developers, which means they have direct commit access to the master git repository. However, they might not be "active" depending on your definition - some have gone a while without a commit. I looked it up and neil moore, MarvinPA, kilobyte, bh, DracoOmega, sgrunt, mumra, and ontoclasm have all pushed commits in the last week, with some community patches too.

Also, who wrote the "philosophy" section of the DCSS manual? Was that DPEG?

Yes, dpeg wrote the philosophy section of the manual.

[rosstin]

It's true that club-butchering situations are no longer a thing... that was the first example that came to me and fulfilled my requirements. With the example I'm trying to give something that essentially illustrates why Crawl is so fun, highlighting tactical and strategic decisions, randomness, excitement, options, variety, emergent gameplay... in as little space as possible. I'm open to suggestions in that regard! If people have really amazing Crawl stories that provide a good example of "what makes Crawl so good?" I'll gladly replace the paragraph.

Aside from the developers who have access to the master Github repository, about how many other developers do you think Crawl has? Let's say in the past 2 months, how many non-official developers have made commits to your other repositories? I've been studying open source development a bit, and it's very interesting to see the roles of various tiers of developers and how they support eachother.

[rosstin]

Here's another one: What lessons has the dev team learned from open source development? What are some mistakes that were made, and what will you do differently in the future as a result? Perhaps something about Mountain Dwarves or Food Reform?

Which dev was the one who was obsessed with 27 of everything?

[Kate]

The Philosophy section of the manual was indeed written by dpeg - I also just remembered that dpeg wrote a bit on the development wiki about open source development: here.

In terms of people who contribute but aren't official developers, those who contribute tiles and vaults for Crawl mostly aren't official developers - they upload to Mantis where someone with commit rights will take a look and add them. Very vague guesses for the past couple of months: 2 or 3 tiles contributors, 3 or 4 vaultmakers, 4 or 5 people who've submitted other patches of some sort (ranging from bug/crash fixes to bigger features like the new Skald spellbook).

[rosstin]

Thanks so much y'all!!

I'm starting to finish up the article. It's far from perfect, but unfortunately my time is finite and I have to finish the article sometime. I'll show you the Gamasutra post as soon as it's finished, and there will be a couple of hours within which it won't be too embarassing to change stuff.

[brendan]

Aside from the developers who have access to the master Github repository, about how many other developers do you think Crawl has? Let's say in the past 2 months, how many non-official developers have made commits to your other repositories?

Here are the number of git changes by author over the past 2 months. It doesn't tell the full picture because commits vary drastically in size and scope.
Members of the devteam (16 active in this period) made 1418 commits, other contributors (17) made 138 commits.

If you want to pull this data yourself, check out our git repo and run ` git log --since="2 months ago" | grep Author | sort | uniq -c | sort -k1 -n -r`

  Code:

    250 Adam Borowski
    244 Pete Hurst
    243 DracoOmega
    188 Neil Moore
    154 Steve Melenchuk
     78 Chris Campbell
     69 David Lawrence Ramsey
     60 lain
     51 Pekka Lampila
     51 Brendan Hickey
     41 Ed Gonzalez
     39 ontoclasm
     19 Samuel Bronson
     14 Florian Diebold
      8 Translators
      8 Raphael Langella
      4 Jason Van
      4 infiniplex
      4 Eino Keskitalo
      3 pubby
      3 Mu
      3 elliptic
      3 Chris Oelmueller
      3 blackcustard
      2 Zannick
      2 Reid Barton
      2 Policarpo Caballero
      1 xFleury
      1 Taylor Hedberg
      1 Rahul Chandra
      1 LexAckson
      1 James Ravn
      1 Alexander Vostres


[rosstin]

It's up! I still have a window to make changes, so feel free to comment, criticize, point out anything I treated unfairly, etc.

http://www.gamasutra.com/blogs/RosstinM ... uelike.php

[brendan]

[url][/url]

However, what [auto-explore] actually did was remove the tedium of wandering around the dungeon in favor of quickly getting the player to the interesting decisions that are the meat of the game: Fight or flee a dangerous monster?

This really gets to the heart of something I've heard from amitp (http://www.redblobgames.com/): Games should compress the boring, mundane parts. A game world with endless forests is immersive, but that isn't where the action happens. When we see something in film, we're probably seeing it because the writer/director believes it to be essential. How many action films show the protagonist using a toothbrush? I'm glad Crawl follows the same tack.

[rosstin]

Thanks Brendan!

I proooobably won't get featured this time... I have a feeling that maybe I should have spent less time explaining what a roguelike is (and their lineage) and more time talking about open-source... but oh well. Always the next article.

The Gama people will probably decide sometime early tomorrow if I get featured. Useful things I might change until then:
* Better takeaways?
* Better images?
* More awesome title?
* A source for numbers/graphs, that show the increase in Crawl's playerbase.

I'll be up early tomorrow morning to implement any last changes before it gets seen by the Gama people.

[evilmike]

If you want some easy stats on the number of players, wins, etc, you could look at the tournament stats. There's one for every majopr release. Here's the one for 0.12: http://crawl.develz.org/wordpress/0-12- ... nt-results

If you look through the backlog of that blog, you can see similar posts for the previous tournaments. These are just for online games, but you can see from this that the game is still getting more players.

There are also some graphs for the online servers. This is hosted on CAO, but as far as I know the graphs count games on all the major online servers. The peaks you see there are the various tournaments. http://crawl.akrasiac.org/scoring/per-day.html

There was also a survey conducted last year which included offline players too, but I don't think the results were ever published.

[Galefury]

Actually some interim results were presented at IRDC2012. See the talk #9 slides here.

[evilmike]

Ah, that's true. I think the survey was open for a couple more months after that though.

[galehar]

What do you think are the key features that set DCSS apart from other roguelikes?

I haven't played many other roguelikes, but I'd say: its size. It's a huge game (but maybe TOME4 is too?). The tight balance (but maybe brogue is even better?). Maybe replayability with the many species, backgrounds and gods.

It's true that club-butchering situations are no longer a thing... that was the first example that came to me and fulfilled my requirements.

crawl has always tried to put the focus on tactics, strategies and fighting monsters. Inventory management, item identification, butchering, hunger, all those things are secondary and keeps getting removed or simplified. Many players would like us to be more drastic with such changes. I have the feeling that lately, this direction has been strengthen.

Who were those original 2 devs? Are they still part of the team?

I think they were Darshan Shaligram and Erik Piper. Darshan still shows up from time to time.

How many devs are active now?

Depends on how you define active, but looking at brendan's stats, I'd say about 10.

Here's another one: What lessons has the dev team learned from open source development? What are some mistakes that were made, and what will you do differently in the future as a result? Perhaps something about Mountain Dwarves or Food Reform?

What I've learned from open source development of a game, is that there's a lot of discussion. It's hard to have everyone agrees on a feature, so there's talk and compromise. And sometimes it leads to nowhere which can be frustrating.
I wasn't much involved in the Mountain Dwarf genocide, but I'd say it was much about a communication problem. We learned something very important from the food reform: simplicity and clarity are essential. Crawl is a complex game, but the complexity emerges from the many interaction of simple features. We had a similar setback with the initial implementation of constriction which was overly complicated, but we were able to simplify it.

Which dev was the one who was obsessed with 27 of everything?

This comes straight from Linley

* A source for numbers/graphs, that show the increase in Crawl's playerbase.

As evilmile pointed out, I'd use the tournament stats for this. You don't talk about tournaments btw, I think it's an important aspect of the game and something which really set it apart from the other roguelikes.

[evilmike]

Regarding 27, my theory was always that the original Rogue had 26 levels, so he made Crawl have 26 + 1. This only applies to the dungeon, though.

[galehar]

Well there's also the fact that 27 is a beautiful number. 3^3.

[jpeg]

Personally, I think that being "open development" (generally open to player suggestions, though all design decisions are made by the team) is actually more important than open source, though the latter allows for players directly contributing to the game. In fact, most devs were recruited from the player base after they'd submitted a number of valuable patches (whether completely new features, bug fixes, or description updates) or at least lots of high-quality vaults or in-depth insight into game design. Because of this, while the dev team has changed a lot (most of the earliest devs are no longer active), development is still going strong. Every new devteam member brings in a unique set of design values and fresh ideas. You're right that the philosophy section certainly helps set a guideline of where development should go. In fact, players are strongly encouraged to submit feedback, vaults, tiles or patches because in the long run that's what makes the game thrive. (The name "Stone Soup", while originally chosen as a bit of a joke, has really come to mean something in that regard.)

I originally got into Stonesoup development because I spent a lot time playing the game, there were things I wanted to improve, and I had lots of free time on my hands. (I dropped out of development when I got a job.)

Also, I've recently starting summarizing the survey results. Might still take a while until I'm finished, though. (So many replies, so many questions, and way too many of them allowing free text answers. Yikes!)

[galehar]

For each new version of Crawl, the devs hold a vote on new features to attempt to implement. The features are ranked and put into tiers, and the dev team races to implement the top-ranked contenders.

Where did you get that? Things have never worked this way. We sometimes query players for popular feature requests, but we have never gone through a formal vote. We even sometimes know that a design decision is going to be unpopular and go with it anyway. Sometimes, players don't know what's good for them. For example, some way to keep buffs permanently active is something many players want. The expectation was also driven by the fact that it was on my todo for some time. But each time we have discussed how we want to implement it, the result was that we didn't want it. Mountain dwarf is another example. The unpopularity of the removal wasn't a surprise, we totally expected it.

While Greensnark and Erik Piper were open-sourcing Crawl

Crawl’s players have shot up since it was open-sourced.

I think Crawl has always been open source. This is what allowed darshan and erik to fork it. What happened is that the official release was stagnant, the maintainer wouldn't commit contributors' patches and wasn't making any significant change either. The DCSS fork has the same license, it's the development model which was more open as jpeg explained.

the developers of Crawl have proven that with the right modeling and philosophies

I think you mean design.

[dpeg]

rosstin: Any decision on the article yet? Would be a shame if it goes to waste -- many thanks for the effort!

Sorry for the late reply, I am on vacation with the kids (and without the mother), so I am basically not around computers.

What I consider most important about DCSS:

• One of the few big roguelikes under active development. (The trend has gone towards games much smaller in scope. This is fine, but I believe it is good to have some deeper & richer games in one corner of the genre.)
• The development model: no money involved; open source; low threshold for players to contribute; low threshold for people to get commit rights (i.e. "become developers") -- on the last one, I have been told by serious people that DCSS is very quick to make that step (which is good).
• The approach to design: always re-assess the current content; not afraid to change/remove content (i.e. not fixated on just adding stuff); obsessed with balance, both local and global; very heavy focus on interface.
• Parts of the theme (this is very personal): Gods (ideally those with actual flavour, especially if it goes against the grain); all other places where Crawl builds its own story (this starts with mascots like Sigmund, Yiuf, Gastronok etc. and includes spells like Malign Gateway).
• I believe that the community (this includes various forums, for example also the one on SomethingAwful, but also ##crawl and the biannual tournaments) helps the game more than is generally given credit for. Webtiles are notable because they allowed a lot more players to participate than before.

Short summary: Crawl is big enough to get away with a somewhat baroque interface but in return gets access to a strategical component and its own universe. Replay value is high and development turnover quite quick for a game this size. Development is open and all kinds of support are encouraged.

I did indeed write the philosophy section and believe that has actually inspired other roguelike developers to explicitly write up where they see their game. These days, I have little to do with the day-to-day development but I am still involved in some godly affairs. (I really hope that some day Crawl will feature all of: randomly generated gods; demigods where the player can actually do something; Lugonu worshipers to desecrate altars to other gods; the gold god).

[njvack]

This is the kind of thing I need to hear, the fact that you wouldn't have played Crawl without auto-explore is important.

I know I'm on the wrong side of history here, but I actually don't like the concept of autoexplore. It kind of says "hey, the exploration part of the game is actually really boring so let's automate that!" It definitely fixes the symptom of "I'm bored pressing keys to explore carefully" but I feel like it kind of skirts the underlying issue -- and that there's maybe a more interesting solution hidden somewhere underneath. Exploration could also, I think, involve interesting choices.

Say, for example, unexplored terrain was somehow a resource for players -- then you need to decide how much terrain to explore and when. Maybe if rest-to-heal (another of my least favorite roguelike tropes) were replaced with explore-to-heal... Or if it had a cost (XP? Piety?) such that fully exploring levels was often not what you wanted to do...

ANYHOW back on topic: one thing that I think really sets crawl apart from other games is the involvement between the dev team and players. This forum (and its activity!) is a rare thing in games in particular. It's pretty awesome.

[mumra]

Say, for example, unexplored terrain was somehow a resource for players -- then you need to decide how much terrain to explore and when. Maybe if rest-to-heal (another of my least favorite roguelike tropes) were replaced with explore-to-heal... Or if it had a cost (XP? Piety?) such that fully exploring levels was often not what you wanted to do...

Desktop Dungeons of course have done this and it works very well for them. Of course in that game, how you manage that resource is a central part of the puzzle. In Crawl I don't think it'd be fun to have to think too much about that, and would actually take some of the fun of exploration away. I do agree that we lose something without exploration, and sometimes my most fun when playing the game are times when, either due to caution or just from getting really immersed for a while, I start manually exploring again (actually I frequently do this when testing layout generators too, since the interesting part is in how spaces lead onto each other). Unfortunately I also got quite used to having autoexplore_delay and travel_delay = -1 during the tournament. These are actually even worse for moving the focus away from exploration; but I find online play too slow otherwise now, and for a game that can be as time-consuming as this, that's quite a big thing.

Anyway; we did actually discuss recharge-by-exploration when going over possibilities for rods, but it seems like this would encourage "stashing" areas of exploration for safe recharges later on, and I don't think that sort of gameplay really fits the ideals of crawl.

If I were designing something from scratch, I'd go with something like: no healing over time, healing only from consumables and spells, MP recharged by collecting balls of glowing blue stuff that sometimes gets dropped when enemies die. But that would be a vastly different game and I don't know if there's anyway to change this in crawl; on the other hand we might have all said that about victory dancing

[rosstin]

Thanks guys! I don't think I got featured this time... I think the article was a bit "out-there" with not enough general interest data. It will be up on the site, but not on the front page. When I have a bit of spare time, I'll try to correct the inaccuracies you mentioned.

[rosstin]

Aw hell! I got featured! Maybe I should spiff it up then.

I changed "vote" to "informal vote". I know the community doesn't actually vote, but the last time I saw a development thread on this forum, the first post was devoted to ranking the likelihood of features to be implemented, and a number of developers and players discussed desired features, and developers had a bit of clout in that if they really wanted something, they would just implement it.

[evilmike]

I remember that thread, it was really just an experiment and I don't think much came out of it.

Almost all actual development talk in this game happens in the ##crawl-dev IRC channel, with more complicated things sometimes going on the mailing list or wiki. Sometimes informal votes do happen, but it's rare. I think it usually happens when discussing whether someone should be given commit rights.

If a feature is especially controversial, it gets discussed a bit first, of course. Sometimes even then it gets committed, but reverted later after some discussion. It's more of an informal consensus.

[rosstin]

Based on that feedback, I altered the wording in that section to be more accurate.

I also changed the semantics of open-development/open-source to be more accurate. I used the word development three times in one line!

[evilmike]

I should also say, there isn't even an informal process of ranking what new features should be added, or planning stuff like that. Except for some very large things, everything is done on an ad hoc basis. What gets added is basically up to what _individual_ developers or contributors feel like working on. Things become more organized only in the month or so before a major release, when things switch into bugfixing mode and certain unfinished things in trunk are likely to get held back. Even then there's no planning involved, just a vague notion that there should be about 2 releases a year.

There's no long term plan, no lead developer, no formal structure for anything. For a lot of projects, this kind of ad-hoc "system" is a bad idea, but it's always worked well for DCSS. It works because of a good dev culture and because discussions happen constantly. Therefore, even though there are no concrete plans, the really active devs tend to know what other people are working on (or want to work on), have chance to talk about design, and so on.