April 04, 2001, 4:09 PM — With version 2.4 of the Linux kernel still hot from the oven, the Linux development community officially kicked off debate about what features should go into the 2.5 version at a two-day workshop last weekend.
On Tuesday the group, headed by Ted Ts'o, an engineer at VA Linux Systems and board member of Linux International, ran down the list of features and capabilities that appeared to be likely candidates for inclusion in the next version.
That list of possibilities included improvements allowing Linux to better support large storage devices as well as large databases; the development of new network device interfaces and Internet protocols; a number of enhanced security features; and better support for Linux working with laptops.
Ts'o said he could not commit to what features and capabilities would be in Version 2.5 since development will not commence until Version 2.4 is completely stabilized. However, he added that enhanced network device interfaces, true asynchronous input-output capabilities, and improvements to the kernel's virtual memory layer are likely to make it in.
"It is hard to say what will make it in, given that 2.5 kernel development has not actually opened yet. We are still in the shakedown period for [Version] 2.4. Linus [Torvalds, creator of Linux] deliberately has waited for a few months on starting a new development cycle so we can stabilize 2.4," Ts'o said.
Ts'o said it is unlikely that there will be dramatic improvements made in 2.5 to Linux's scalability in terms of its SMP (symmetrical multiprocessing) support. Many developers attending the two-day workshop felt the improvements made in 2.4 will satisfy the needs of enterprises at the lower end of the computing spectrum for now.
"SMP scalability is much better in 2.4 than 2.2, although there is more work to be done there. But we scale very well on eight [processors] and up to 16 chips," Ts'o said. "We did discuss the prospect of SMP clusters, but that work right now is more [research and development] in nature," Ts'o said.
Likewise, the long-awaited Journaling File System was not a major focus of the workshop, and Ts'o declined to say if that technology would be supported in Version 2.5.
Given that Version 2.4 took about two years to complete -- which Ts'o attributed to the overly ambitious development goals of the project -- developers discussed ways of speeding up the development cycle for Version 2.5.
While no specific approaches were agreed on, Ts'o said one approach would be to stabilize a certain number of core capabilities earlier in the process and then port them back to the existing version.
"Maybe what we could do is once we make substantial improvements over 2.4, we can immediately start to stabilize those improvements. With some subsystems we could also think about porting them back into 2.4 for people who really need them now," Ts'o said.
