Interview: Keith Packard

Keith packard is a long-time X11 hacker, and currently works at Intel. He is the project leader to X.org.

What's the goal of your talk at FOSDEM 2007?

I'm hoping to give current X developers some exposure to the larger community as well as encouraging new developers to come join in the fun.

You have been at FOSDEM before, right?

Yes, I've been to FOSDEM several times before and always find the conference stimulating and fun.

So do we :-)
Do you plan to get some hacking done during the conference?

I'm hoping to spend most of my time meeting with other developers and enjoying their company.

How far off is the future where X.org requires no configuration at all? What about hot-plugging of devices?

X.org 7.2 has made some improvements in this area already, with the ability to actually use EDID data from monitors to detect preferred modes. The RandR 1.2 work has improved on that so that we can automatically configure multiple monitors with sensible defaults. On the input side, the Xinput hot-plug work has resulted in a restructuring of the input system so that HAL can now manage the input devices used by the X server.

Most systems should now be able to function without any configuration file at all, the largest issue at this point is that the presence of a configuration file disables much of the automatic mechanisms.

With the new Xinput work and RandR 1.2, we'll be able to hot-plug input devices and monitors. What we're still missing is the ability to hot-plug video cards; that looks to be a significant challenge given the current X server architecture.

And beyond that work -- how will X.org evolve in the future?

Of course, the real answer is 'in whatever way people choose to change it', and I am only one of many people interested in what changes still remain necessary. For my own part, I hope to help release a new version of the server in a few months incorporating some new acceleration code along with the RandR 1.2 changes. After that, I still need to figure out how to deal with mouse coordinate transformations and polygon rasterization issues.

Polygons, tha't 3D-related then. How do you feel about the introduction of accelerated OpenGL to window managers?

I think it shows off the capabilities of the system in new and exciting ways. While we continue to encourage most applications to use the basic 2D drawing operations to ensure the widest possible portability, the addition of 3D effects to the desktop offers an opportunity to explore new desktop management styles while providing people with an eye to 3D graphics a better outlet than screen savers to show their work.

Do you think that in the future, an even larger part of the graphics processing could be accelerated in hardware?

At this point, the core X and GL APIs can be entirely accelerated by hardware. We've got some work to do in moving up the stack, especially related to media presentation, including full H.264 encoding and even video display improvements of the sort currently implemented entirely with the CPU by free software media players.

Considering the graphics drivers... Do you think that Intel's positive attitude toward open source drivers could influence other manufacturers as well?

I certainly hope so. The overall X architecture improves best when developers working on multiple platforms come together to develop solutions. We work very hard to avoid creating standards based on a single architecture, and having several key manufacturers not fully engaged in this process often makes development more difficult.

One of my key reasons for joining Intel was to continue to push for better support for free software from Intel and to use that to try and encourage other vendors to do the same.

Thank you for these insights.

Thanks for inviting me back to Fosdem this year; I'm looking forward to another fun weekend in Brussels.

Additional links:

This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.

FOSDEM

Links