The sudokant is back. It took Yahoo's Slurp many months to exhaust its backlog of moves to explore, even getting 404s on every try. My wife remained a faithful puzzle solver and it was her complaining about this puzzle, due to USA Today, that led me to check to see if it was safe for sudokant to reappear. -- WardCunningham

* http://c2.com/~ward/sudokant.cgi?..5...1...1..759....21.3...2.37.4..8.5..2..4.4..3.87.6...4.78....453..9...7...2..

[''To solve that puzzle efficiently, work out the 4th, 5th and 8th numbers of the second column.'']

----

The page name says everything.

These:
* http://c2.com/~ward/sudokant.cgi
* http://c2.com/~ward/sudokant.cgi?7.8...5.....2.1...3.........4.3.........8.1...26....93.9.......6...7......45.....
seem no longer to work.

Is the source available?

See also:
* IsTheBlankSudokuDiabolical
* SuDoku

''Copy now recovered - thank you.''
----
'''Discussion:'''

''Perhaps it has been rekanted? It wasn't exactly correct in its results. It should be able to find at least one correct solution to solvable puzzles and distinguish unsolvable puzzles.''

That was never its purpose. It was intended to do much of the mechanical work, not to actually solve the puzzles.

''As I understood it, its purpose was to solve the blank sudoku using a minimal set of rules (ideally without backtracking). If its intention is to "do much of the mechanical work", however, it fails even that... it certainly shouldn't be filling blanks with wrong numbers. Yet it does. You're turning the human into a system that does the mechanical work of undoing what Sudokant just did... hardly a profitable venture.''

You obviously never used it.

''Not on a regular basis, no. And I wouldn't reccomend it any more than I'd reccomend Microsoft's little paperclip program. Both have provided entertaining toys for a few minutes, though, before their level of error and weak interface became annoying. I take it you're a regular user? and you find identifying and cleaning up Sudokant's mistakes is more... efficient and/or satisfying... than simply filling the blocks yourself? In my experience, finding and eliminating errors after a bad number was written down is a great deal more difficult than filling in blanks with the correct numbers in the first place. What I'd want is a program that can, when I hover over a square, can display 2-4 alternatives if the number of possibilities has been reduced sufficiently (i.e. so I do not need to write numbers in boxes then cross them off). THAT would be a useful reduction in mechanical work.''

You wrote:
* "... their level of error ..."
What error? Sudokant only ever showed what possibilities were left in each cell. It couldn't make errors, since it made no choices. I never spent time "... identifying and cleaning up Sudokant's mistakes ..." because it never made any.

You seem to be talking about a completely different program.

''I must have been using it in a different manner. It is the same program.''

* How then were you using it? The commenter above is correct in pointing out that Sudokant was incapable of error.
** Are you going to reply?
* ''Actually, Sudokant produces a spelling error in its descriptive text every time it's used.''
----
Ward's code applies two basic rules: ChrisGarrod suggests a third rule
	1. Each row, column and square must contain each digit at most once.
	2. Each row, column and square must contain each digit at least once.
	3. The solution must be unique. 
20071104 I never did figure out who the participants in the above conversation were. I did discover a third rule which Ward's code could incorporate, but I've no idea how. I'll try to paste in some examples soon. I've used that rule a few times in real puzzles when I've been unable to proceed towards a solution.

p.s. Ward, did your logs show any robot that found a solution? I put stuff in my robots.txt as soon as you said they were trying.	-- ChrisGarrod