The Big Kolab Kontact Interview - Part I

KDE Dot News recently spoke with some prominent people from the Kontact and Kolab projects. We talked about how both projects got started and how they have evolved. Enjoy the first part of this two-part interview.

First of all can you explain who you are and what is your role in the Kolab/Kontact project?

Tobias Koenig: My name is Tobias Koenig, I have been doing KDE development for 5 years and since 3 years ago I've been the maintainer of KAddressBook.

Cornelius Schumacher: My name is Cornelius Schumacher. I started my KDE career by contributing to KOrganizer and I have maintained it now for something like five years. At aKademy I passed on KOrganizer maintainership to Reinhold Kainhofer. I also do various other things for KDE, one of them is trying to take care of kdepim generally.

Steffen Hansen: My name is Steffen Hansen, I work for Klarävdalens Datakonsult AB and have been on the KDE project more or less since the beginning. I am the lead developer of the Kolab server.

Bernhard Reiter: I am a managing director of Intevation GmbH, which is a Free Software company. For Intevation I coordinated the Kroupware and Proko2 contracts.

What are Kolab and Kontact, and how do they relate to each other?

Steffen Hansen: Kolab is a Free software groupware solution. The components are the Kolab server and Kontact, which is the KDE Kolab client. There is also a Kolab web client in the works.

Tobias Koenig: Kontact is part of KDE and integrates the main KDE PIM applications (KMail, KOrganizer, KAddressBook, KNotes) into one GUI. So it has the appearance of MS Outlook/Evolution but the advantage that all components can still run as standalone applications. Kontact supports access to various groupware servers and one of them is the Kolab server.

Bernhard Reiter: Kolab consists of several components:

• Kolab Solution Architecture
• Kolab Server implementation
• Kolab Client implementations (especially the KDE Kolab Client)

So I would say the current Kontact can act as KDE Kolab Client.

How did the Kolab project start?

Tobias Koenig: Martin Konold, a KDE developer, gave a talk on LinuxTag 2002 in Karlsruhe, which described a groupware solution built upon Open Source Software that can be used as replacement for Outlook/Exchange.

Bernhard Reiter: I have to comment on Tobias here. The Kolab Server was not designed to replace MS Exchange. We addressed this question in our Kroupware FAQ:

"So no, it's not a goal to build a replacement for Exchange or Outlook. On the software side we have assembled a solution that solves some typical asynchronous groupware tasks like group calendaring. Our usage of standards like disconnected IMAP and the iCalender format is a new, innovative approach."

What exactly does a Kolab server do? Because sharing folders on IMAP is not a new concept.

Steffen Hansen: Correct! The server uses well-known and available free software components like OpenLDAP, Postfix, Apache and Cyrus IMAP server plus some custom code to glue it all together.

The features you get from the Kolab server in addition to just using the individual components (mail, web, imap,...) are:

a) Central configuration. The tweakable configuration parameters, user database and so on are stored in one central place: The LDAP database. Centralised administration of the Kolab server is done through a web-based interface.

b) Multilocation support. Because of a) it is possible to set up multiple Kolab servers that share configuration and user database. It works by the use of LDAP replication. I see this as a very nice feature for both supporting multiple weakly connected locations and for being able to scale up to multiple servers to support huge numbers of users.

c) Groupware specific functionality. The Kolab server contains code that does automatic appointment handling for group and resource accounts on the server. It also creates and manages freebusy lists for all accounts.

I heard about Kroupware and Proko2. What's with the names? How do these projects relate to each other?

Tobias Koenig: The first project was called 'Kroupware', it consisted of Kolab1 and a KDE client, which was mainly a modified KOrganizer with embedded KMail.

Steffen Hansen: 'Proko2' (which stands for "Project for Kolab2") had a main focus of enhancing usability of the client and adding features to the server to allow groups of people to collaborate, do server based virus scanning, the multilocation feature and so on.

Bernhard Reiter: The names refer to two contracts to deliver a groupware solution. Our companies did this with Free Software and started Kolab. Kolab is a Free Software project and thus has potentially different players and goals than what the contracts might say, so we needed to come up with names to decouple this and being able to openly talk about it. For example we did not want to overpower the naming of the client done by the KDE community.

How many companies are involved with the Kolab project? I see the names Klarävdalens Datakonsult AB, Erfrakon PartG and Intevation GmbH passing by. Can you eleborate a bit on your cooperation?

Steffen Hansen: These three companies combined their efforts to create a crossplatform groupware solution. But we need to mention a few others as well:

Radley Network Technologies is a South African company that make a plugin called Toltec Connector for Microsoft Outlook. It allows Outlook to use IMAP for mail and groupware data. Radley's has put a big effort in adding the Kolab-specific features to the Toltec Connector to allow it to interoperate with the Kolab server and other Kolab clients. Among other things, this includes support for reading and writing the XML format that Kolab clients use for internal storage of groupware data. Communication between clients is done with iCal/vCard.

A fifth company also deserves mention here: CodeFusion. Like Radley's, they are also based in South Africa. CodeFusion offers professional Kolab support, they have helped out with implementing some server-features and they also work on the web client.

Microsoft Outlook uses the legacy TNEF format which is not capable of working properly with iCal. How did you guys find a solution for that?

Steffen Hansen: Well, let's first make a distinction between the format used for communication and the format used for storage. Both Kolab1 and 2 do indeed use standard iCal for inter-client communication (ie. sending invitations etc.). Kolab1 also used iCal as the storage format to store your calendar etc. on the imap server. Unfortunately is has not been possible to create a solution that allows Kontact and Microsoft Outlook to share the same storage based on iCal. iCal is too flexible in some respects and not flexible enough in others. Therefore Kolab2 uses an easy to read/write and well documented XML format for the internal storage.

Creating our own format has been frowned upon by some, but I really think that it has been a big help for us to have our own format that we could tailor to our specific needs to be able to make both the KDE client and Outlook 'happy'. For the end user, the visible result of using this new format is that he/she can switch back and forth between Outlook, Kontact and the web client without having to import and export data between them.

Citadel/UX, a BBS system claims to be Kolab1 compatible these days. Does this mean we can use Kolab compatible clients like Kontact with Citadel?

Bernhard Reiter: As far as I remember they are not fully compatible with Kolab1 because free/busy handling, directory server architecure and lack of legacy support make it slightly different. Apart from that you probably can use it with Kolab1 clients.

Cornelius Schumacher: Citadel/UX is quite interesting because it has a long history and is a mature server system. There once was a KDE client for it which used KOrganizer as the calendar component. I would love to see some more interaction between the Kontact and the Citadel/UX developers.

One problem we have with groupware servers is that there are no really accepted standards for them. There are several drafts for standards and quasi standards set by implementations, but there is no generally accepted standard for access to groupware servers. It would be great if we could work into the direction of something like a common way to access groupware servers. We did make some very small steps but there is still a long way to go. I expect that this will become more important in the future as the current situation isn't really beneficial for anyone.

Kolab stores mail folders using the Cyrus "maildir" format. Citadel stores everything in Berkeley DB. It seems you are using IMAP to access files on the server, which uses the filesystem and returns you that file. Tell me.. why didn't you use a database engine for this like Citadel/UX does?

Steffen Hansen: We strongly believe that you need a storage mechanism that fits the problem to be able to scale up. Mail messages are variable sized independent pieces of data stored in a tree-like folder structure. This looks a lot like a filesystem to me. For an example, look at the statistics for Carnegie Mellon (the home of cyrus imapd):

I seriously doubt that you can scale that high with a database-based solution.

Using mail messages for storing calendar information might seem odd at first, but it is actually quite nifty. Having one calendar entry per mail makes it easy to merge data when you have manipulated your calendar from several places. Also, conflict detection/resolution can be done nicely with this approach.

A lot of people complain about the installation procedure of Kolab. Has the situation improved? What made it such a pain to install anyway? Was it because of the OpenPKG format?

Tobias Koenig: For Kolab1 the installation procedure was really a pain but this has been improved in newer versions. Kolab2 Server is installable by simply executing one shell script. The reason was not the OpenPGK format, it's used for Kolab2 as well and has a lot of advantages. I guess the main reason was the limited time frame and the developers didn't put much resources into implementing an installer but tried to get a stable Kolab version finished.

Steffen Hansen: Now the server installation has been streamlined a lot. After fetching the install script, all you need to do is this:

./obmtool kolab
 [wait while it builds and installs]
 
/kolab/etc/kolab/kolab_bootstrap -b
 [follow on screen instructions for initial setup]

/kolab/bin/openpkg rc all start
 [now the Kolab server is running. You can go to the 
  webgui and create users and so on]

Will Kolab in the near future be able to integrate more with an enterprise authentication system for single-sign-on? (for example OpenLDAP, Samba with LDAP support and maybe if you have Kerberos/Heimdal set up as well).

Steffen Hansen: I can not make any promises, but this is something that has been on our own wishlist for some time.

Will Kolab have a mailman interface one day? Any plans for that?

Tobias Koenig: It's not planned yet. Nevertheless it's possible from the technical point of view, so volunteers are already welcome :)

People on the MS Windows platform can use the Kolab server through the use of the Toltec connector. What would be a compelling reason to use Kolab instead of let's say MS Exchange?

Steffen Hansen: Using a Kolab server instead of MS Exchange buys you better scalability, higher reliability and a soft migration path away from proprietary software on the desktop (because you can gradually introduce Kontact and it will share data with Outlook).

Tobias Koenig: Kolab is fast! Kolab2 will have the same functionality as Exchange, and it's Free Software.

Another very interesting topic is the cooperation with the Free Software project KDE. How did that work out? How were decisions being made with regard to implementations?

Tobias Koenig: The cooperation went quite well, most of the staff from the company group come from the Open Source community, so you can talk and work with them like with every other Open Source developer. There were also three meetings, organized by Intevation GmbH, where KDE-PIM developers and people from Klarävdalens Datakonsult AB attend to discuss the further steps for integrating Kolab support into the official KDE release. Some evidence of the success of these meetings is that KDE 3.3 was released with full Kolab2 support.

Cornelius Schumacher: I'm speaking from the KDE perspective and have to say that it was an interesting challenge. I really appreciate that the Kolab project explicitely includes integration of the client code into KDE development. In practice there have been some problems because the Kolab people actually didn't have enough time to do this integration in a smooth way, but in the end, I think, we can be happy that the Kolab client has finally arrived in KDE.

It's a general problem of how to handle integration of third-party code into KDE as there are a couple of different interests which have to be balanced. Commercial contributors usually produce a lot of code which might be hard to properly review for Free Software maintainers doing this job in their spare time and which is developed with a different goal in mind from the code coming from inside of the project. The most important point is to talk with each other as early and as much as possible. There have been a couple of personal meetings between Kolab and KDE people and they have shown to be extremely valuable.