kolabd package available in Debian sid

Andreas Gungl Andreas.Gungl at osp-dd.de
Thu May 11 09:43:10 CEST 2006


Hi,

I've dealt a bit with Horde some time ago to make it basically work with 
Kolab 2. Some things got fixed, some have remained open. Anyway, Horde is 
in a usable shape for our installation (in-house, ~70 users, mix of Toltec 
and Kontact clients, web client only used when accessing from remote 
sites).
Nevertheless there are still problems from my point of view, one of them has 
been named (the database dependency). I'll comment on them below.

Am Mittwoch, 10. Mai 2006 19:42 schrieb Jan Schneider:
> > On Sat, May 06, 2006 at 03:33:59PM +0200, Richard Bos wrote:
> > What my understanding of the design problems is:
> >
> > 	a) the use of the database versus Kolab's design of imap folders.
> >            It will create almost unsolvable sync issues and hurt
> > stability. And if the code is spread throughout
> >            Horde, this will be hard to change.
>
> Horde does not have an inherent database driven design. Quite the
> contrary, it has a driver based design. This made it possible to write
> Kolab drivers for the several applications so quickly. There is this
> DataTree thingy that still causes a lot of headaches because it's the
> last remaining database dependency in Horde, but solutions for this
> problem have already been discussed in the past.

The database dependency is ugly. At first, you have to install MySQL on a 
machine only for a small thing which has nothing to do with the actual 
functionality of the server. (I'm aware that I could use a second machine 
for Horde besides the Kolab server, but that's something else but a 
consolidation, right?)
Additionally, I've seen several accounts messed up by manipulations in the 
datatree made by Horde. I could resolve the problems by manually deleting 
the datatree contents for those accounts. But that's all but funny.

> > 	b) Horde needs higher privileges. This is in contrast
> > 	   to the Kolab-Webinterface that only works with the
> >            priviledges of one user. If true, this would hurt the
> > security. Note that database application are often designed this way,
> > Kolab can do differently which is a plus.
>
> I'm not sure if this is true, and I'm wondering if this is really a
> general design issue in that case. I guess it could be solved.

It is at least a security flaw. It would be better to let the user log in 
using his own credentials (checked against LDAP). A related problem was the 
ability to log in both with the Kolab UID and primary email address. There 
has been an attempt to fix that, but I don't know about the result. I've 
adviced our colleagues to use the primary email address, because otherwise 
they may get unpredictable results when accessing their data.

> I'm not trying to sell anything, but I really doubt that it makes any
> sense to write a custom web frontend for Kolab, even if it ties much
> better into the server. Horde already provides all features that the
> Kolab server offers, without installing a Kolab server. I would be
> insane to start from scratch instead of ironing the few remaining
> wrinkles out.

What makes me wonder is how few interest has been shown from the Horde team 
when I submitted tickets and patches. There still is a good chance for 
Horde to become THE web based client for Kolab - with a lot of publicity. 
Well, at least even you, Jan, are reading this list. I see it as an 
improvement. But it's still not what I understand as constructive 
cooperation between projects. AFAICS Horde still can't use the personal 
contacts from a Kolab account. As another example, 
http://bugs.horde.org/ticket/?id=1659 is still open although there should 
be no problem to fix it for an experienced PHP developer (unlike me). So, 
yes, I'm disappointed and I understand if others are as well.

As has been said, we're using Horde despite the flaws. I've decided to keep 
my hands off of the Horde files. ATM we still run a CVS copy from about one 
year ago (June 1st, 2005). Perhaps I'll try a newer release if somebody can 
give me good reasons. (We'll certainly switch to the latest release as soon 
as we make it to Kolab 2.1 which might take a while though.)

Best regards,
Andreas




More information about the users mailing list