+1

[Kenneth Marshall]

On Wed, Feb 04, 2009 at 01:50:23PM +1100, Bron Gondwana wrote:
> On Tue, Feb 03, 2009 at 05:27:28PM -0500, Wesley Craig wrote:
> > On 02 Feb 2009, at 01:32, Bron Gondwana wrote:
> >> It does change the regex syntax in sieve, and is probably non- 
> >> standardly evil! )
> >
> > Oh?  As I (minimally) understand sieve and it's extensions, the regex  
> > syntax is thoroughly specified.  Deviating from those standards is  
> > probably a problem.
> 
> Yeah, probably.
> 
>     Implementations MUST support extended regular expressions (EREs) as
>     defined by [POSIX.2].  Any regular expression not defined by
>     [POSIX.2], including [POSIX.2] basic regular expressions, word
>     boundaries and backreferences are not supported by this extension.
> 
> Ken wrote the spec :)  Guess we should harass him if we want the
> canonical interpretation.  Still, our users seem happy enough with
> the PCRE setup we have, even if it's not entirely to spec.
> 
> > Also, perhaps if no length-oriented regex interface is available, the  
> > body extension should be disabled.  I can easily imagine body searches 
> > causing a denial of service, otherwise.
> 
> Yeah well, if and when it becomes a problem I'd consider it.  Besides,
> Linux and all the BSDs support a length-oriented interface.  Solaris
> doesn't seem to out of the box, but then I don't know my way around
> it well enough to be sure.
> 
> There's a bunch of other libraries to choose from anyway, RX, TRE,
> something will do the trick.
> 
> Bron ( I assume you can shoehorn glibc onto Solaris if you want too )
> 

PCRE provides a much nicer environment for regular expressions
and I think that they meet the spirit of the spec admirably. We
use them here in a number of software products and the perl
regex syntax is much easier to work with.

My two cents,
Ken