Re: Text collation

[Carl Witty]

On Wed, 2006-12-06 at 07:41 +1100, Kevin Ryde wrote:
> Carl Witty <address@hidden> writes:
> >
> > If I understand correctly, this would make date->julian-day (for
> > instance) convert from Julian calendar dates before 1582 and Gregorian
> > calendar dates after 1582.
> 
> Yes.
> 
> > This is cute, but I don't think it's particularly useful;
> 
> Apart from being what people actually used then!
> 
> > the problem is that the Julian to Gregorian switch
> > lasted for hundreds of years around the world (and, indeed, some
> > churches still use the Julian calendar to determine religious dates), so
> > you can't tell if a date is Julian or Gregorian just by looking at the
> > year.
> 
> Yes, the switch happened at different times, but if you're calling it
> gregorian then it makes some sense to go with the date pope gregory
> signed off on :).  It could get some localization perhaps, though that
> sounds like hard work.  I think the gnu "cal" program does that.

I believe that virtually any program that needs to do date computations
on historical dates will need more support than SRFI-19 provides, no
matter what meaning we assign to "date"s with years before 1582.  If you
consider the small set of Guile users who need to do date computation
for historical dates, and the even smaller set for which "before 1582 is
Julian, after 1582 is Gregorian" is exactly the date semantics they
want, I think it is quite possible that this smaller set is empty: that
nobody would ever correctly use the SRFI-19 "date" type for historical
dates.  (My totally-made-up estimate is that the likelihood of wanting a
split in October 1582  is about the same as the likelihood of wanting
the proleptic Gregorian calendar, which is at least used for ISO 8601;
and that both of these are far more likely than wanting a split which is
automatically computed based on the locale of the user running the
program.)

If it is indeed true that nobody would ever correctly use a
split-at-1582 date type, then the decision should be made on the basis
of the difficulty of implementing, testing, and documenting.  It seems
like implementing and testing the proleptic Gregorian calendar is far
easier than the split-at-1582 calendar, and the difficulty of
documenting either is about even.

> > I would propose leaving the code the way it is now, and changing
> > the note to something like "Note: Dates in this module use the proleptic
> > Gregorian calendar,
> 
> If the srfi author wanted to clarify that then it'd be ok, otherwise
> surely a calendar with at least some sort of historical basis is more
> use.

I note that the SRFI does say "A date is a representation of a point in
time in the Gregorian calendar..."

Carl Witty