I agree with you but ... if data has no clean header (which is the simplest task in data-collection) then I suspect some of the rows will not be cleaned either. For example, there might be missing fields in rows, truncated or overflowed fields (because of some weird instrument error or network error or conversion error) or formally correct fields but qualitatively wrong (I mean an uncalibrated instrument etc.). So, if (yet another) outlier detection module comes out of this, provided nysus has the time, then let it be, I say.