Out Roving Rove. I have argued below that any possible arguably legitimate purpose of Rove's leak to Cooper could have been achieved equally well if Rove had not identified the career CIA employee who, he alleged, proposed Wilson for the mission to Niger as Wilson's wife. Now I realise that the illegitimate purpose of smearing Wilson could have been achieved as well (that's not saying much) if not much much better without identifying that employee as Wilson's wife.

Rove could have told Cooper "Wilson was just sent to Niger because of a recommendation by a career CIA employee who has been sleeping with him for years."

Now that would be a real Catch 22. It would tend to suggest that Wilson was having an extra marital affair (and not the gender of that CIA employee is not specified mmmmm).
Of course the sleeping arrangements of Plame and Wilson are strictly intramarital, but Wilson couldn't point that out without blowing Plame's cover. I mean what is he supposed to say "it wasn't a woman I have been sleeping with it was my wife" ?

Now I am reasonably confident that "a person who has been sleeping with Joe Wilson for years" is a unique identifier (and if not it is none of my business) but the Rober Luskin like violation of Gricean attribution would throw foreign intelligence services off the scent of a woman who happens to be married to Joseph Wilson.

Such a statement by Rove would be neither a crime nor a lie. I guess that's why he didn't say it.

nonupdate post script: Gricean attribution theory (explained to me by my friend, the spooky genius, Ehud Reiter) is a rule for guessing things which are not literally implied by the definitions of words combined with logic. It is a major challenge for computer scientists who like major challenges (like Ehud) to program computers so that they can tell why Luskin is weird. The idea, If I recall correctly and understood in the first place, is that there are conventional ways of phrasing things. These conventions are useful in understanding what people say, because they are normally violated only when they must be. For example (Ehud's example) if someone says "I saw an animal on your front yard" it is normal to assume that this animal is not a dog. The reason is that, if it is possible to use the word "dog" not the word "animal" then people normally use the word "dog". The simplest level of Gricean attribution thery is that there are words which are used a lot (first order words or something) and other words which are only used when there is not (first order word or something) which fits. The favored words are typically very short (one syllable) and, of course, often used. The theory goes on and on. Basically conventions for how we phrase things which can be phrased in different ways are useful, because it is possible to interpret things people say using the rule that, if it were possible to say something with the same meaning a more conventional way, the speaker would have done so. Thus we can assume that the intended meaning of a statement is not one which could have been stated in a simpler or more conventional or preferred way.

Robert Luskin freaks people (including me) out, because he says things which are technically true but misleading. They are misleading because the correct interpretation of his statements has a meaning which could be expressed more simply (in a more conventional way, in a preferred way).

The use of "a CIA employee who has been sleeping with Joe Wilson for years" for "A CIA employee who is married to Joe Wilson" for "Velerie Plame" uses such weirdness for a patriotic sliming purpose. Patriotic, because foreign intelligence service employees will have difficulty fighting the assumption that this CIA employee is not Wilson's wife. Slimy because it insinuates that Wilson if fooling around in a way which he can not refute without blowing his wife's cover and breaking the law.