(looks like Alexey hit the "reply" button rather than "reply all".)
On Sun, 2005-10-02 at 21:18 +0100, Alexey Melnikov wrote:
Kjetil Torgrim Homme wrote:
On Tue, 2005-09-27 at 13:31 -0700, Ned Freed wrote:
I'm in favor of the behavior being consistent across all tests. If it
isn't
we now hove to specify this obscure bit of behavior for every future
extension that introduces a test. Not good IMO.
I agree.
I agree with Ned's statement, however this doesn't answer my question:
do you want to always strip leading/trailing spaces for all tests or not?
I do not particularily like the stripping, and ideally I think [ACAP]
should have specified that i;ascii-numeric ignores leading whitespace.
one alternative way back then would have been to register a new
comparator which handles true numbers (ie. negative numbers and possibly
even decimals) where leading whitespace is disregarded.
however, can we change [RELATIONAL] in such a backwards compatibility
breaking manner?
I see it as a question of which implementations do we break, and since there is
some chance that others have implemented these specifications in places we've
never heard of the bias needs to be towards keeping the status quo.
The reports I saw seemed to indicate that most implementations followed the old
specification in this regard. Our didn't, but it wasn't hard to change it to
comply. I also note that we implemented part of this but not all, so either way
this goes we'll need to change things.
My position is therefore that we should keep things the way they are. This
isn't the most convenient answer for our implementation, but given the
information we have available I think it is right one.
Ned