First: Today, even LWN's comments are "worldwide publications" so... well... And that's the problem raised by the good old central database and the Internet as it is today. No way to have something like a random conversation somewhere that fully fades away in a short time (or other minor things that do not deserve posterity at all). That's something new with respect to privacy. Well, not so new now that it's 20-years old; but honestly I still find it difficult to adapt to the new scheme.
Second: Your reasoning is an interesting example of the potential impact of information inference, frequently forgotten.
From what I state, you can also deduce further things; and you also have to reason about the truth of what I say. And me, I would have to take into account when I write what every reader can deduce (no offense, but well... especially for Mary ;-).
And in this case, you did not even include information from multiple sources for your deductions (employer and job position for example).
What seems to me is that, confronted with these new issues in a connected world, there is nearly no way to fight against this privacy invasion; except by using the same tools to build more trust as a compensation.
For example: I'd like my computer to tell me more about readers of this comment that also look at my profile on a profesional social network, or that look at my physical address, etc. I don't think I would invade their privacy more than they would mine, I would be protecting myself as well as establishing a better trust network *with* them.
Very soon, we will rejoin: identifying liars is key to the topic.
Not only in front of a judge (though in this case, it is also legally reprehensible). Like investigators, one usually need evidence to spot liars and make a difference between random guys and... trustable husbands. ;-)