Alan Cox Interview

LWN

Alan Cox is a name that should be well known to most Linux users. He is responsible for much of the code in the Linux kernel, including small pieces like networking and SMP. He has handled the latter-day 2.0 releases, and serves as one of the primary conduits for patches into the current development kernels. He is certainly one of the key players in the kernel development arena.

We thank Alan for taking time out of his busy schedule to answer a few questions.

LWN: 2.2.0 has been a long time in coming. Too long, perhaps? Do you think the 2.3 series will be shorter?

AC: I hope so - I'd like to see 2.4 being out early 2000 but it depends entirely on Linus. I do think there is a general desire for a much shorter turn around time for 2.4.

LWN: What are the most interesting things that will happen in the 2.3 series?

AC: I'm not actually sure. It's possible to predict some things by looking at patches that are pending post 2.2 or in progress. The clear ones are 64bit file support, probably 32bit uid support and support for very large numbers of processes. On the device side, I'm expecting I2O and USB to be in 2.3. I'd hope the PCI bus/PCMCIA/Cardbus/Hotswap clean ups that need to be done occur. That actually has scope to reduce kernel complexity rather than increase it, which is a good sign.

I'm not sure a lot will happen in the networking arena that is visible to end users - addition of DECnet isn't like to be a major win to the average user. ATM might well be more visible and it is about time it got in.

I'm looking forward to all the palmtop support being folded in - there are several strands here and all of them are beginning to show clear unified needs.

LWN: You seem to have taken on an increasingly organizational role in the kernel development process. Does that sit well with you, or would you rather get back to more full-time hacking?

AC: It just sort of happened. I'm anticipating the amount of co-ordination will go down again now that 2.2 is basically sorted out. I hope so anyway as I want to get the I2O code finished, debug the 3c527 ethernet driver I am writing and use the work the vMac people have done to attempt to get palette loading and floppy disks working on the Macintosh 68K.

LWN: And how do you manage to process (and produce) all that mail every day and still get something done?

AC: That is actually fairly easy. I've always been a fast reader and most of the stuff I dig through requires little in the way of a reply - a lot of the mail I send is just pointing people at each other to avoid duplication of work. Much of the rest is tracking bugs and patches.

All the patch merging and bug chasing is getting something done, so the mail scanning is definitely not time wasted.

LWN: Kernel development over the past year has been interrupted occasionally by "Linus burnout" episodes. What are your thoughts on the sustainability of the current development model? Is there anything you would change?

AC: If I was the kernel organiser, there are quite a few things I'd do differently. Right now Linus applies all the patches and builds the trees, I'd much rather there were a group of people directly merging patches into the kernel tree and Linus sitting watching it and vetoing things rather than doing all the merge work, too.

The model has changed over 2.1.x and it has evolved to a kind of compromise that seems to work very well. Linus is still applying all the patches but there are people now collating and feeding Linus tested sets of patches in small cleanly organised groups. Larry McVoy's new version control toys may solve some of the remaining problems.

LWN: Do you anticipate taking on responsibility for the 2.2 series like you have with 2.0, or will somebody else have to step up for that one?

AC: I never really anticipated it with 2.0.x; it simply happened because I was collating all the patches. My guess is that, by the time 2.2 gets into that long term maintenance state, 2.0.x will be basically dead.

LWN: If you were to point at the biggest unmet need in the Linux world, the project most in need of volunteers currently, which would that be?

AC: That's always a hard question to answer. One problem is that, as Fred Brooks observed, adding man power to a late software project can actually make it later. This is true with free software as well.

Clear areas that could do with more work are better GUI tools for the kernel facilities. A nice graphical ISA PnP manager is one example. Others include some end user friendly tools for the new 2.2 bandwidth control and management functions.

The free software world doesn't really, however, work like a managed corporate structure. If someone is going to do something as a volunteer, it has to be something they find fun. Watch freshmeat and also for calls from people like the FSF (eg they are currently after more documentation people) and see if something tickles your fancy.

LWN: What do you think was the most significant event in the Linux world in 1998? Any idea what will be the most interesting development of 1999?

AC: The most visible one was clearly the sudden discovery of Linux by the suits and very much tied to the Mozilla event. I'm not sure what surprises 1999 will hold. Microsoft's current attempts to commit corporate suicide are bound to have some effect on the Linux world in 1999. Whether they will be the most significant is hard to tell. The other one will be if large PC vendors start to make machines with Linux or no OS available. The recent fun with the Windows Refund saga will undoubtedly help this. It's actually important that it becomes easy to buy a machine without an OS. It would be bad for people like the FreeBSD community if most of the people fighting for OS choice simply said "OK, now you can have Linux" and left it at that.

My guess is 2000/2001 will be when the really big stuff happens. That I suspect is the time scale for big Unix vendors to begin openly switching to Linux. For vendors whose revenue stream is primarily support and hardware, the math is simple enough.

LWN: Along those lines, what are your thoughts on the future of the BSD variants? Will Linux be their undoing? Is our relationship with the BSD systems what it should be?

AC: I don't think Linux will kill FreeBSD. I can see one of Open or NetBSD dying. At one point, I'd assumed NetBSD was doomed but it has a very definitely stayed alive.

In the longer term, I expect that Linux will help them. Supporting the Linux kernel API (something they already do fairly well) will give them the same application base that Linux is creating. With SCO, BSDI and apparently Solaris going to support the Linux kernel API, we should see a lot of applications for Linux running just fine on anyone's favourite OS.

LWN: How do you feel about the increasing corporate interest in Linux? Does Linux risk "losing its soul" as some people fear?

AC: Linux has always reflected its user base so I'm sure that some parts of it will turn more corporate. I don't actually see it as a big problem. No large corporation can "own" Linux or take away the right to freely distribute and change it.

Personally I don't mind if someone releases a Linux distribution aimed totally at corporate IT managers. I'm sure the technical community will use words like "boring, out of date, slow to change" about such a distribution. I've met corporate IT managers - words like "boring and slow to change" have them excited.

Linux already has this spectrum - from the corporate style Caldera use, through the "easy to use/install" of Red Hat and the "pure and free" vision of Debian. It's richer for it now; I don't see why it should be poorer for it in future.

LWN: We all have much fun reading Telsa's diary. Any hints for others who want to be hard-core hackers and stay married too? And how did the battle of the cuddly penguins end up, anyway?

AC: There is a difference between being a hacker and the "socially inadequate computer geek" vision of the press. It's definitely true that there is a lot of overlap. A lot of good hackers are, however, members of the human race and know what is going on in the real world.

The thing that I suspect matters most is that Telsa is more important to me than sitting in front of a computer reading email.

As to the penguins, they are currently all sitting on top of the printer paper in a sort of group hug. Now that I have a Wai Yip penguin to compare with the Linuxmall and other penguins, I should probably write a cuddly penguin review.

Copyright 1998