Personal Bytes

So yesterday was another Open Source Session as part of the Google Tech Talks. The host was the well known australian Andrew Morton, who’s the second line maintainer of the Linux kernel and also responsible for the -mm development tree (basically, anything that reached the -linus tree went through -mm first).

This time the presentation was held in Building 43, just in the middle of Google Campus. I was completely awe-amazed. My jaw kept falling down. Before commenting on the presentation, some interesting stuff:

• First search query I saw when I arrived at the lobby: fat chicks
• Most of the Google employees that I saw on that building were running Ubuntu
• One guy was working with 4x Dell 22″ TFT displays at the same time
• There was a mockup of a plain inside the building sketched by Paul G. Allen
• Free food all over the place
• Google rocks

So, back to the main topic. The State of the Linux kernel! Andrew Morton is not a formidable speaker. He does fine but he’s not a Steve Jobs. On the other hand, he does talk about utterly useful content and does not distract you with side issues. A lot of content to talk about then! Here’s the list of the work we should expect to be done in the Linux kernel for the months to come:

Server

• Infiniband
• Network protocols, Congestion Management
• SATA/SCSI evolution
• NUMA evolution
• Virtualization (KVM, VMWare, lguest, Xen)
• Containerization
• Resource Management
• kexec and kdump
• kproves and systemmap
• ext4
• Large SMP and Multi-core enhancements
• Hotplugging: devices, CPU, nodes, memory

Desktop

• Ongoing power management work
• Neverending stream of framebuffer drivers
• Direct-rendering drivers
• Much work ongoing with Input, Sound (mainly Novell/SuSE taking care of ALSA), USB, 1394, DVD Subsystems
• Interactivity improvements (Disk scheduling, Dirty memory writeout, CPU scheduling)
• File-change notifications

Consumer/Embedded

• Dynamic ticks, hi-res timers (needed by OLPC)
• Ongoing memory footprint reduction
• Improving NOMMU support
• New architectures (FRU, awr32, blackfiw)
• OMAP, SPI

The feeling I got is that they’re headed to where Linux clients want them to be headed. That is good. We, consumers, make the rules, not the other way around.

On a side note, it was easy to notice that every single change, regarding whether it comes from a single contributor or from IBM, Novell or Redhat is still filtered by the Benevolent Dictator (Linus) so things haven’t changed that much and considering the evolution of the Linux kernel I’m quit glad they haven’t. It works!

Now some random notes. Someone asked whether we could expect other languages to be used besides C (and ASM), like C++. Andrew would like so but convincing Linus is another issue. Also, someone talked about reiserfs3, 4 and what’s going to happen. According to Andrew, reiserfs3 is pretty much stalled and nothing should be expected out of it. As for reiserfs4, patches are still coming from the Namesys guys but the pace of development is slowing down considerably. XFS and ZFS were also brought to discussion but Andrew said that until XFS changes the way it saves data (ie, it saves the meta-data first meaning that a crash before the actual data being saved would fill the file with useless zeros) we shouldn’t expect it to become mainstream.

And that was it. Quite interesting, a lot of good food and nice people there. The tech talk should be available online in the very near future.

P.S. - For those who don’t know, Andrew Morton is currently a Google employee.