Yes, when you submit a comment to HN you intend for it to be published on the HN site. However, using comments for text mining purposes is a different thing and in principle consent is required to do that. The default case is that user submitted content remains copyrighted by the author, with an implied consent for the site to publish the material on the site. There can be an explicit license such as with Wikipedia or StackOverflow; in that case text mining may be explicitly allowed.

It would be a godsend to the #hnwatch crowd looking for sentences to take out of context.

Perhaps not in the content. But I'm not so sure there might not be actual legal issues (in jurisdiction that have privacy/data protection laws) with regards to stuff like being able to see patterns in login-times (based on post meta data), and similar. It's tenuous -- but just because things are a) on-line in the form of the main hn site, and b) accessible through the api, doesn't automatically mean that giving raw access to the data doesn't carry with it implications for whether or not the data should be considered personal/(more)sensitive.

I'm assuming this data set offers something the API doesn't (that's the point of uploading it, right? Other than just being an easy way for others to "back up" the hn data) -- and that can very well mean there are privacy/personal data implications.

I'm mostly stating this because of the principle, not because I think it's anything wrong with this concrete upload of hn data (and wrong != legal/illegal anyway).

As an example, at least in Norway, while you may have two separate databases that are linked to people, such as a database of cars passing through toll booths, and credit card charges from a large bar/restaurant chain -- it would generally be illegal to link the two.