> I'm not sure how a date in the format YYYYMMDD can have an epoch of 1601.
IBM mainframe Cobol contains builtin functions which convert between YYYYMMDD strings and integers - and 1601 is the epoch they use for that integer representation. I assume the person you were replying to was talking about this fact, just stating it somewhat confusingly
> The epoch for a year YYYY is 1* BCE (since there is no year 0).
Well, there is a year 0 in astronomical year numbering. One can say that 0 CE = 1 BCE and -1 CE = 2 BCE and more generally n BCE = -(n-1) CE for all integers n > 0. Maybe “0 CE” is incorrect per a strict definition of “CE”, but it is correct if we define it less strictly
> * You're not really supposed to use dates < 1582 in ISO 8601 though without prior agreement though it's meaning isn't really defined.
There’s really only two choices: proleptic Gregorian, or Julian. I don’t know why ISO 8601 doesn’t just mandate proleptic Gregorian with astronomical year numbering. But I suppose in the rare cases a computer system needs to represent pre-1582 dates (such as historiography), Julian is the norm. I suppose they could extend the syntax to specify which calendar is being used
> Edit: There do exist standards that default to 1875-05-20 when there's a null date however (GIS-related), but I've not seen anything that suggests the SSA uses it.
Do you know which ones specifically? Would be interested to know this