How bad a boss is Linus Torvalds?

ComputerWorld ^ | 11 November 2015 | Steven J. Vaughan-Nichols

[ShadowAce]

I've known Linus Torvalds, Linux's inventor, for over 20 years. We're not chums, but we like each other.

Lately, Torvalds has been getting a lot of flack for his management style. Linus doesn't suffer fools gladly. He has one way of judging people in his business of developing the Linux kernel: How good is your code?

Nothing else matters. As Torvalds said earlier this year at the Linux.conf.au Conference, "I'm not a nice person, and I don't care about you. I care about the technology and the kernel-that's what's important to me."

Now, I can deal with that kind of person. If you can't, you should avoid the Linux kernel community, where you'll find a lot of this kind of meritocratic thinking. Which is not to say that I think everything in Linuxland is hunky-dory and should be impervious to calls for change. A meritocracy I can live with; a bastion of male dominance where women are subjected to scorn and disrespect is a problem.

That's why I see the recent brouhaha about Torvalds' management style--or more accurately, his total indifference to the personal side of management-as nothing more than standard operating procedure in the world of software development. And at the same time, I see another instance that has come to light as evidence of a need for things to really change.

The first situation arose with the release of Linux 4.3, when Torvalds used the Linux Kernel Mailing List to tear into a developer who had inserted some networking code that Torvalds thought was--well, let's say "crappy." "[A]nd it generates [crappy] code. It looks bad, and there's no reason for it." He goes on in this vein for quite a while...

Here's the thing, though. He's right. I read the code.

[ShadowAce]

[who_would_fardels_bear]

In 2011 the Linux "kernel" was 15,000,000 lines of code.

I remember back when the Unix kernel was truly a kernel.

15,000,000+ lines of code is 15,000,000+ opportunities for hacking exploits.

Linux is becoming as bloated as Windows.

[VanDeKoik]

Feminists have been trying to get at him, from what I understand, to the point where he absolutely refuses to be alone with any female at anytime for fear that someone will try to lodge a “sexual assault” complaint at him.

[so_real]

So, this :

   mtu -= hlen + sizeof(struct frag_hdr);

which could have been improved like this :

   if (mtu < hlen + sizeof(struct frag_hdr) + 8)
      goto fail_toobig;
   mtu -= hlen + sizeof(struct frag_hdr);

was instead re-written like this :

if (overflow_usub(mtu, hlen + sizeof(struct frag_hdr), &mtu) || mtu <= 7)
   goto fail_toobig;

I think Torvalds' rant was entirely justified, and that makes him a "good boss". A "bad boss" would have looked the other way and allowed that crap to take root and grow.

[Bogey78O]

Agreed.

“Tough but fair” used to be the ideal not the malady.

[Steely Tom]

Magic numbers aside, I’m confused about how adding a “goto” to C++ code is an “improvement.”

[miss marmelstein]

Tough doesn’t mean saying “I’m not a nice person” to your employees. Plus, I’m sure his employees knew that without him “explicating” it.

[ShadowAce]

There's nothing wrong with goto if it is coded properly.

Using it just to create spaghetti is not coding properly. Using it to exit is, IMHO.

[ShadowAce]

He doesn’t have any employees.

[Moltke]

a bastion of male dominance where women are subjected to scorn and disrespect is a problem

LOL. As if these geeks even have a chance of getting within 10 feet of a female...

[Steely Tom]

There is never any need for “goto.”

I’ve been writing C++ code for twenty years, and have never once used “goto.”

[ShadowAce]

Embedded code within a limited hardware platform?

[glorgau]

Looking at the example given, I agree. Why be “nice” about it. Blunt saves time and effort.

I HATE people that write that kind of garbage to make themselves look smart and also to make themselves the only ones that can maintain the code.

[Steely Tom]

C++ is not a good language for embedded platforms.

The overhead required for object-oriented features such as polymorphism and inheritance is not justified in a limited-resource environment.

[ShadowAce]

I coded for over 10 years without using GOTO while using COBOL in a high-volume financial mainframe environment.

Then I moved to a small company coding for ATM machines, used C ( a little C++ ), and discovered a need for goto. One has to pack a LOT of functionality into a very small coding space. Sometimes it required GOTOs to save code.

[uglybiker]

How would you act if your folks named you Linus?

[Steely Tom]

One has to pack a LOT of functionality into a very small coding space. Sometimes it required GOTOs to save code.

I agree, but I thought this thread concerned Linus Torvalds and the Linux kernel, which who_would_fardels_bear stated was comprised of more than 15 million lines of code. That is by definition not going to be packed into a "very small coding space."

I don't have anything against GOTOs in very resource-limited environments.

It's been a long time since I had to pack anything into 64K of memory.

BTW, I once wrote code for the RCA 1802, and debugged it with an oscilloscope. Back when Body Heat was brand new, and Kathleen Turner was super hot.

[ShadowAce]

...this thread concerned Linus Torvalds and the Linux kernel, which who_would_fardels_bear stated was comprised of more than 15 million lines of code.

Yeah--that's a lot of code.

My first question, is how much is used in any one installation? A lot of that code is determination for hardware platform. I'd actually like to see multiple versions of the kernel, based on the hardware it's aimed at.

Just separating 32-bit from 64-bit code would shrink it a lot. I know that it does depend on the compiler, but there's quite a bit of code in there regarding that as well.

Also, compiler optimizations and code optimization would be a huge help for shrinking the memory footprint and speed.

[sten]

i would imagine the vast majority of the 15m lines would be drivers