Andrew Morton, the Linux kernel v2.6 maintainer at the Open Source Development Labs, spoke with InfoWorld Contributing Editor Paul Venezia.
IW: The v2.4 kernel release was viewed as a watershed event for Linux and brought significant improvements for server-class systems. What can we expect from v2.6?
AM: 2.6 will scale significantly further than 2.4 on large machines. More CPUs, more memory, more disks, larger disks, more threads, etc. Also, it is smoother and more responsive on the desktop. The merge of uCLinux is nice for very small embedded systems.
IW: Which features of the v2.6 kernel do you consider to be the most integral to enabling wider adoption of Linux for large-enterprise uses?
AM: The scalability. And some code device management infrastructure, which should permit more sophisticated device identification and naming facilities on large systems.
IW: Many Linux administrators feel that the v2.4 kernel really "jelled" near v2.4.10, and wasn't really ready for production use prior to that release. Might this happen again with v2.6?
AM: I hope and expect not. The 2.5/2.6 stream has been stable and usable for over a year. I haven't run 2.4 on any of my machines since 2.5.15 or thereabouts. The testing that has gone into 2.6.0 is much more complete and more successful than that which went into 2.4.0. I'd say that 2.6.0 is at roughly the same level of stability as 2.4.17 or thereabouts.
IW: What immediate changes or improvements do you foresee for v2.6?
AM: Not a lot, I hope -- getting the main kernel synced up with the various external trees (the different CPU architectures tend to get out of date) and mopping up various things that were missed or not fully fixed for the 2.6.0release. After that, I expect we will have a period of integrating work that the various vendors have been maintaining outside the 2.4 tree. And the regular flow of new drivers.
IW: What do you see as the next stage of Linux kernel development? What might v2.8 bring?
AM: Not a lot has been talked about at this stage. One … is adding `virtual server' support. Maybe clustered file system support as well. Most people are still concentrating on 2.6 work; I expect that more will be known about the 2.8 content after the July 2004 kernel summit.