A year ago I've completed my first "scientific" paper. I presented it in a local conference for junior researches. It actually didn't have much effect: as far as I understood no one at the conference was into the subject (ah, no, I'm wrong, there was one guy who developed boost::build). I wanted to fix it (since it contained minor bugs) for a very long time. But since I haven't done it, then, I suppose, I will never manage to fix it.
Small abstract: it aims to describe why C++ libraries may lose binary compatibility. It also analyzes if certain compatibility restrictions may be relinquished if the user of the library is only allowed to use a subset of C++ by library's specifications. No one did such analysis before.
The paper is also a somewhat successful attempt to analyze C++ with scientific method. It appeared that this language is not eligible for such analysis, since it's just too huge. But anyway this is the most complete--and thus the most useless--compatibility guide. It was quoted by GCC folks in their manual, but that's an only known reference; and I have no idea how they found my article.
The developer of ABI Compliance Checker used the knowledge gained during this study, and this checker happens to be a working tool. This proves that my studies haven't been useless or erroneous.
Well, on this weekend I'm committing another article to a conference; and I hope it will be less useless to the world. Wish me luck.