This is a position similar to a ten thousand year ko. White takes the ko, but after Black passes White should not fill.
passes and
fills.
almost fills with the bulky five shape. White is not obliged to capture however, so
passes and
captures.
White almost fills with the bulky five. passes and
eventually captures. This time, the eye shape is sufficient to life unconditionally.
Therefore, neither side should fill the ko because they end up worse off than just leaving it open. They would both like to capture the single stone in the ko, however, and this could create endless repetition. Spight rules end the game after White takes the ko for the second time, because Black passes for a second time in the same position. The game stops "unnaturally" favouring Black.
Bill: Very interesting position. :) BTW, at the end I think you mean, favouring White. If indeed, neither player wants to fill the ko -- and I think that is so --, then this challenges all rule sets except forms of no pass go. For instance, under no pass go with prisoner return Black has no prisoner, and so may as well resign after White takes the ko.
To conclude, it is undesirable for either side to fill the ko under area scoring. If Black fills the ko, Black would lose 6 points (from B+2 to W+4). If White fills the ko, White would lose 2 points (from W+2 to draw). It can be critical, e.g. if komi is -1 point. Under Kee Rules, such situation is an example which yields "no result" if both sides leave the ko unfilled after winning the ko.
(Additional note on 13-Jun-2010: In the latest version of Kee Rules, instead of no result, this example would lead to a win of White. It is fair to argue that Black should be penalized because Black is the last player to create such initial board position.)
I suddenly come up with an idea - why are we just focusing on the ko prohibition but never the scoring system? If area scoring system is slightly modified so that any stone groups in the ending board position with only one liberty is not counted, then such situation, no matter who is occupying the ko, would yield draw anyway. In other words, such ko fight does not make any difference in value unless really filled. In general I believe such modification can solve many anomalies as well. Bill, what do you think?