The original 4chan code was in Japanese and moot used Babelfish to try figure out what did what. From what I remember, the original Futaba code is just as bad (http://www.2chan.net/script/). It's no secret that 4chan is cobbled together with glue and string, moot has said this several times before.
The comments are in Japanese but if you know PHP you should be able to understand it without the comments anyway. It's a single file of less than 1K lines and written in a rather straightforward style. It's also extremely easy to setup - just edit the config parameters at the top and drop it on a webserver, and it's ready to use. Not even a database is needed.
I don't think it's "bad"; it's damn simple and works well for what it is, and contains no unnecessary complexity. No dependencies on some other huge framework, multilayered overengineering, or excessive generality. The same can't be said of the many other clones of it that were attempting to be "better designed" rewrites.
As for "maintainability" or all that other software engineering stuff: this board script doesn't really need to be maintained, because it works, and if anything needs to be changed, it's so simple that the changes can be made easily. Along the same ideas as http://suckless.org/philosophy
4chan has grown to be quite obviously more sophisticated than Futaba, which really has stayed pretty much the same for over a decade. You can see for yourself:
It's kind of surprising though. I wonder what the results of a 4chan re-write by 4chan users would look like. Or even splitting out some functions for re-writing.
