I'm wondering if someone should be starting a new generation of TAOCP as wiki. As wonderful as TAOCP is, frankly, first volumes are bit outdated. Many problems that were mentioned opened have had huge progress and many new fields have opened up since their publications. Several Level 50 problems are now being taught in CS Masters or some even Bachelor levels. Besides I really find MIX too arcane and getting in the way. In a way, it's kind of disturbing that we still fall back on ~40 years old material for the field as fast moving as Computer Science. Sure, much of the CS is timeless but we are ignoring so much progress and growth in fields such as Computational Geometry, Number Theory, Cryptography, Semi-Numeral algos and so on. Also, even if Knuth manages to complete all of his planned volumes, it would still have scratched only tip of the tip of a iceberg. Of course, some of material will start becoming stale pretty much as soon as it was created. Currently Wikipedia is not a great option to create such writing because it doesn't allow content in "teaching style" or things like creating exercise. If we can distribute this effort out with editor that can approximate Knuth's meticulousness, it could be great boon for our field. Knuth as an original flame bearer could be a great person himself to start such an effort.
There is Wikibooks for this kind of content, but I suspect that compiling and editing a book like this (hundreds of pages of content, extensive exercises, references to the literature, proofs and equations, etc.) would be very difficult on a wiki.
You need contributors who can dedicate enormous amounts of time, a good wiki system (most can't cope with automatic cross-references to equations or sections, and don't provide special formatting or indexing for exercises, proofs, etc.), and editors who can combine the efforts of the contributors and produce something coherent.
I'd like to see software that can handle this kind of text. ScholarlyMarkdown[0] is trying to do it, and something programmable like Pollen[1] would be easily extended to support TeX-like features.
The software won't solve the time sink problem, though. Knuth is like a monk, sitting in his room pushing out manuscript pages. Who else is willing to dedicate the time?
This is a big enough project to justify a bespoke wiki-style system. Perhaps a thesis project for someone, along with tooling to convert existing content from TAOCP. Get some kind of grant to cover employing CS-literate undergrads on a UROP project to do the more difficult data entry and QA.
Heh, nothing's stopping you from starting that wiki. If you can get through any of those volumes, as 'outdated' as some of the information may be, you will be far ahead in algorithm analysis skills than 99.9999999% of the programmers out there. The key is focusing on any volume instead of being distracted by the new findings in the field. I tell this to myself too many times but don't listen. Sigh.
The lone genius working for years on a book is as far from a wiki as can possibly be.
Creating a quality computer science wiki with quality editing seems like a fine thing to do but "TAOCP as wiki" is just a contradiction in terms. You will invite certain comparisons that make you look foolish from the beginning.
If you want to work on solving the problems you're describing, one of the surest paths would be to translate the existing MIX code from the older books to MMIX versions.
I would love to contribute to a new, more open TAOCP that strives to cover a wider range. But I think it will be difficult to organize this.
