ietf-mta-filters
[Top] [All Lists]

Re: Sieve base-spec revision I-D

2005-02-22 13:23:10

On Tue, 2005-02-22 at 13:26 -0500, Mark E. Mallett wrote:
On Wed, Feb 16, 2005 at 05:36:26AM +0100, Kjetil Torgrim Homme wrote:
I wish
there were an escape for including octets by value in strings, e.g.
"\x00".  that would be an incompatible change, can we do that?

(if so, I'd want "\uXXXX" and "\UXXXXXXXX" for Unicode, too (same syntax
as Java and C99))

I agree with wanting to be able to specify octets by value.  I'd hate to
do it with a leading alpha character though, as '\x' is supposed to mean
'x'.  I've done it a la C escapes, with '\0nnn' meaning an octal value,
where you only have to make '0' special, and I don't think \0 sequences
being special would come as as much of a shock as \<alpha> sequences
being special.

perhaps, but I don't think it makes much difference.  FWIW, I did a
quick survey of our user's Sieve scripts (a little more than 2000
scripts), and none of them had extraneous backslashes.  only one user
had (wrongly) written "\." in a regex context.

I don't like \0nnn since it is AFAIK a syntax not used in any other
language, also octal is a bit old-school :-)

(C-style \nnn for octal is problematic since it is variable width.)

I could also see a \0<alpha> sequence having extra
meaning, e.g. '\0xnn' for hex.  

What would be the need for \uXXXX that a series of multiple escaped
bytes couldn't provide?

it's more convenient than having to do the UTF-8 coding yourself.  not
that it's hard, I could even do it in my head, but it's error prone.  it
also puts an additional burden on Sieve script builders, they need to
convert back and forth before presenting the rule for the user.

-- 
Kjetil T.


<Prev in Thread] Current Thread [Next in Thread>