[Kroupware] Re: Preannounce: Kroupware Project started

[Hamish Rodda]

On Thu, 12 Sep 2002 04:03 am, Marc Mutz wrote:
> On Wednesday 11 September 2002 17:04, Hamish Rodda wrote:
> <snip>
>
> > The uploading part is taken care of with timsieved, which implements
> > their "managesieve" protocol.  I wrote a kioslave for it, which I
> > still have if anyone would like it. I also remember seeing something
> > similar in kdenonbeta, though at a glance I can't see it right now.
>
> The one in kdenonbeta was written by Simon Josefsson, well-known for his
> managesieve/sieve mode for emacs and his involvement in the Sieve WG.

Yep... looking at the two they're quite similar, both made from the kdevelop 
template it would seem :)

> It currently lacks the ability to return the sieve capabilities of the
> server in meta data. Are you interested in implementing that in either
> slave?

Yep, I've dug out my old code and made the modification - it sets the 
capabilities as metadata on the result of any stat().  I'm also working on 
the last few features that managesieve supports.  Where should I put it when 
it's done? kdenonbeta... maybe as kio_managesieve?

> Another thing would be to write a common abstraction layer above
> tcpslavebase for IMAP-like protocols. It could handle SASL mechanisms
> and provide the low-level parsing for the IMAP/ACAP/managesieve data
> structures (strings, parenthized lists,...).

Sounds nice, and I wouldn't mind porting my slave to it once it's done, but it 
sounds like a bit too much work for me right now.

> > To me the time consuming part is writing the gui to create the sieve
> > scripts in the first place.
>
> I'm taking care of that for 3.2 in KMail.

Excellent :)

> > > The common thing with these two is that it's done in the client,
> > > and that is the best. Consider a server with a million users
> > > spending 99% cpu time on filtering - not exactly impressive. So for
> > > now we won't spend time on it, but it's an open project, so if
> > > someone volunteers...
> >
> > Do you have numbers to back that up?  From appearances, sieve seems
> > designed to handle this load... "The language is not Turing-complete:
> > it provides no way to write a loop or a function and variables are
> > not provided."  With only the particularly savvy users likely to
> > write very complicated scripts, I'm pretty certain you'd be in the
> > clear.
>
> <snip>
>
> Q> require [ "regex", "body" ];
> Q> if body :regex "(.*)*" { ... }
>
> can be pretty nasty. ;-)

I see... still, this is hopefully something only advanced users would do, and 
they could be discouraged with friendly advice in the GUI, or by disabling 
the regex module.

Cheers,
Hamish.