More Marcelo Tosatti

Frank writes: "There's an interview over at developerWorks Linux Zone with Linus's latest lieutenant Marcelo Tosatti. He talks about what it takes to be the maintainer of the Linux kernel, what his plans are for 2.4 and his favorite hack." If you missed it, you may also want to visit the answers Marcelo gave to Slashdot readers.

Comments

Gotta represent (er, maintain)

[Skyshadow]

Does it strike anyone else as strange that the Linux kernel is still run by a small monarchy?

Seriously: Linus is the king, and he's surrounded by a small contingency of advisors who filter what gets through to him. I'm not suggesting that these people aren't all very deserving, but it seems odd that nobody else is cranking out any sort of alternative. MS or Sun can't be considered serious competitors (not on the same page), and all the BSD's seem to have been pushed to the fringe. This leaves other Linux kernels, and there are none.

I suspect this is because you just can't compete with Linus -- after all, he is the man. Still, it seems to me that this leads to a lack of internal competition in a very important area of overall systems development, which can't be a Good Thing (tm); consider how much KDE and GNOME have benefitted from having each other to race against. The kernel, on the other hand, exists mainly on the preferences of a small number of people.

Of course, Linus historically has shown great insticts; he's only been really wrong once that I can remember. This might sound like a call for fragmentation, but I still can't help but think that being open is good, but being open and competing against someone else is even better.

Re:Gotta represent (er, maintain)

[metacell]

There may be a simple practical reason that such a small group of people maintains the kernel:

It takes a lot of work.

Anybody can program kernel code (or at least try :) in their spare time, there is no time limit, so you can just use the time you have over, and do as much or as little as you want.
But maintaining the kernel is nearly a full-time job. It puts everything else aside.

Or what do you think?

Re:Gotta represent (er, maintain)

[ukryule]

Does it strike anyone else as strange that the Linux kernel is still run by a small monarchy?

More a meritocracy than monarchy (i.e. if you prove you're good enough you get the power) - in this case Marcello proved that he was technically competent and so was given the responsibility of maintaining the kernel.

Of course the difficulty with any meritocracy is who decides your "merit". Unsurprisingly in this case (as in most cases) it's the ones in power (i.e. Linus & Alan). Thus whether it is a true meritocracy or not depends on the abilities of those leaders to pick out the best contributors ...

Still, it seems to me that this leads to a lack of internal competition in a very important area of overall systems development, which can't be a Good Thing (tm); consider how much KDE and GNOME

Surely you're arguing for external competition? In which case, that nice Mr. Gates seems to putting up a decent fight. And internally there are several branches of the kernel floating around, and the major Linux companies often seem to bundle their own version of the kernel.

However, I agree with your central point that Linux does still rely heavily on one man. What happens when he stops running the show is an interesting question ...

Competition is there I think..

[stray]

I'd say the competition is right there, between the contributors. For one problem, various people might submit different patches and solutions, and eventually one of those is chosen to make it into the kernel. Arguably, that is a kind of competition.

Right, it's not driven by market forces but decisions are made by the kernel maintainers, but still... I'd say a lot of people use patches (especially driver stuff) before they make it into the kernel tree, so there's a certain amount of democratic feedback going on abour what patch might be the best for a task or a problem.

Competition in the Kernel.

[metacell]

I think there isn't any competition between different Linux kernels, simply because it's not needed. The current team is doing such a good job, and it's easier and better to make an addition to the kernel than to split the source tree.

Or are we talking about the user's need to choose between different kernels?
Well, you can run roughly the same software on the FreeBSD kernel as you can on Linux. Gnome, Konqueror, Ghostscript... it's all there.
There aren't many commercial vendors selling BSD versions, but that doesn't matter so much when you can just get BSD and install the software you need yourself.

But if there was a significant need for alternative Linux kernels, I'm sure the competition would crop up faster than you can say "ego-boosting Linux fanatic". ;-)

Re:Marcelo

[DA-MAN]

Oh come on, give him a break. So far the only mistake was the 2.4.17rc3/2.4.18 error. He's a human, we all make mistakes. He's doing a good job and things are stabilizing at a good rate. If you don't like it, submit a patch. Make a difference, don't talk shit. Or if you're not technically capable of that, how about writing up a Linux Kernel HOW-TO. Come up with a process, if it's good it may end up actually being used. He seems to be a guy pretty open to anything.

Good move.

[sydneyfong]

Oh, that Slashdot interview! I didn't like it. I was doing a hundred interviews a day, so I was like, "Aagh, no more interviews!" and I answered their questions very fast, and people got angry because of that.

I guess the developerworks guys were smart enough to have the interview done at this time. Just imagine those "I'll work hard to maintain the kernel" answers they'll get if they didn't wait for a few months before they did the interview.

Why doesn't Linux do what everyone else does?

[Anonymous Coward]

If you send in a Perl bug report and used the perlbug utility, nobody has to ask you configuration questions because they have the answer. If you didn't use it, then they just say to send 'perl -V'.

With OpenBSD they use dmesg in the same way as 'perl -V'.

Why with Linux would they have to go back and ask questions? Isn't configuration information (detected hardware etc) available somewhere? Why not just have a utility that sends it in attached to your bug report?