About LWN.net

A talk with KDE developer Kurt Granroth

Liz: Sorry to ask the same question that you've probably heard from so much of the press this week, but could you give me your reaction to the GNOME announcements this week?

Kurt: Actually, the question we've been asked most this week is "When will KDE 2.0 be available?" We have some great things coming out with the release of 2.0, particularly Konqueror, vastly improved management of file extensions and great performance. People are really excited about it. They love KDE 1.0 and KDE 2.0 is a tremendous improvement on that. The conference is going very well for us.

We're not concerned with the GNOME Foundation announcements; it won't change any of our plans for the future.

Liz: Do you see KDE ever creating something like the GNOME Foundation for KDE?

Kurt: No, if we did, we'd probably lose half of our developers. It is very important to us that the developers maintain total control of the plans and directions for KDE. We would not tolerate the creation of any kind of oversight committee.

Liz: In talking with one of the vendors that was part of the GNOME press conference, I asked him why his company chose GNOME over KDE. One of his answers was the energy of the GNOME community. There is also an active and energetic KDE community. Do you think that it is being underrepresented and overlooked?

Kurt: Yes, I do.

Liz: Do you plan any actions or activities to try to change that perception?

Kurt: We've discussed it but not made any commitments. We would like people to be more aware of just how active our development is and the contributions that our community is making [seriously paraphrased].

Liz: The other reason mentioned for choosing GNOME was their use of Corba. Didn't KDE work with Corba at one point?

Kurt: Yes, we did. The initial development on the KDE 2.0 branch was all done with Corba, using the Mico ORBb.

Liz: What changed that?

Kurt: There were several reasons. We worked on the KDE 2.0 development starting when we were still stabilizing KDE 1.0. We spent months on it but were unhappy with the results. The performance was very bad. That, we knew, was potentially fixable through the use of a different ORB, possible Orbit. In addition, though, the complexity of the code, the complexity of the Corba standard, was such that the code became more and more difficult to work with. Eventually, we only had a small group of seven or eight developers who really understood it and could work with it. That was a huge bottleneck for development. Then one of our developers sat down and developed KParts, an alternative to CORBA, within only a short amount of time. He showed it to us. Not only was it blindingly fast, but the code itself resembled the code we'd already written for KDE 1.0. That made it possible for all of our developers to work with it immediately.

Then we had our KDE developer meeting in Europe. We sat down and looked at the potential cost of dumping CORBA. We would be losing the ability to use remote components, but we didn't know anyone currently using that ability and felt it could be added later if people really wanted it. Interoperability with other CORBA implementations wasn't a major impact; everyone builds their own layer on top of CORBA (like Bonobo) and you really have to be compatible with that, not just CORBA, in order to interoperate. So all we really felt we lost was the buzzword. That isn't important to us as developers, so we moved to KParts.

Liz: Thank you, Kurt.