Abel Braaksma wrote:


The default should be the same in both cases, I will fix this.

  

The difference between these two defaults corresponds to my findings, 
indeed. My personal favor goes to the same language in all cases, 
instead of any system default, whether or not the Java api has a 
mapping for it. That way, my xslt will run the same regardless the 
location you run it, even when using defaults for the parameters of 
the format-date function. Of course, this removes the added 'feature' 
of having a processor that 'adopts' to its surrounding when it comes 
to locales (but I hardly consider that a feature and would rather want 
to set the default using a processor-dependent property, which, if I 
may add, is a recommendation that is also mentioned in the xslt specs, 
same section).




My mistake of interpretation. About the default, the spec is indeed 
clear, rereading shows me this line, meaning, imo, that the function 
should return the same value in either of the two cases:



"Calling the two-argument form of each of the three functions is 
equivalent to calling the five-argument form with each of the last three 
arguments set to an empty sequence."



About what I call 'recommendation' (what's in a word), I am sure I read 
it some place, but I cannot find it anymore...



-- Abel Braaksma