I suggest to remove the parentheses around "later"
> (listElement",")* (later)
I also wonder whether there are some reasons for not having longYear in
the lists. I suggest to add them (read below).
There are some differences between the BNF's lists and the lists as of EDTF spec.
The BNF accepts lists with mixed consecutives, such as
whereas the EDTF spec do not give any example of that. I can't find use
cases for such lists, but if there are use cases for such lists, I would
suggest to just add such an example in the EDTF spec to make it obvious
for the reader that EDTF allows it and then everything is fine.
If there are no use cases, I would suggest to modify the BNF, which could be done strightforwardly as
> consecutives = date ".." date
consecutives = ("-")? ( yearMonthDay ".." yearMonthDay | yearMonth ".." yearMonth | year ".." year | century ".." century | longYear ".." longYear) | "-" ( yearMonthDay "..-" yearMonthDay | yearMonth "..-" yearMonth | year "..-" year | century "..-" century | longYear "..-" longYear )
where I also added the longYear for consecutives.
Another problem is the fact that the BNF accepts
which should be avoided. To avoid lists with only one term is easy, but on
the other hand, it should maybe (or maybe not) also accept
In other words, there is a need for a few deeper changes in the BNF as a whole, when it comes to "x" and "u".
Of course, we could take "the easy way" and just discard both
and choose to rewrite the last two ones as
as a non-list without braces. But this wakes the question of the semantics
of "One of a set" lists vs. "u". What's your opinion? If no use cases are
available, I would suggest to just discard single-element lists in the
These questions are of course related to the thread on "proposed direction
for uncertain and approximate".