Summary: Tracing the source of the claims about Linux being "bloated" leads to no surprises and the allegations are soon being refuted

THOSE in the press who are constantly Linux hostile have become preoccupied with Torvalds' admission that Linux is getting larger (more code). Is it getting more "bloated"? That's debatable.

Little attention was paid to the original source of the claims. It is not as though Linus Torvalds came up with the statement. He was more or less being fed words in the sense that he was merely presented with a sort of complaint, which he was then left to acknowledge or deny based on no independent evidence. Torvalds does not perform benchmarks, so we at Boycott Novell contacted those who do. We spoke to Phoronix and others who are more intimately familiar with Linux development (kernel space).

First of all, a question worth asking is, where did the claims come from? It turns out that a Novell employee, James Bottomley, is the one who raised the issue. We presented a video of his in this post about Microsoft's (and Novell's) increasing influence inside the Linux Foundation. Here is a report from The Register, which was cited very widely:

During a roundtable discussion at LinuxCon in Portland, Oregon this afternoon, moderator and Novell distinguished engineer James Bottomley asked Tovalds whether Linux kernel features were being released too fast, before the kernel is stabilized.

Citing an internal Intel study that tracked kernel releases, Bottomley said Linux performance had dropped about two per centage points at every release, for a cumulative drop of about 12 per cent over the last ten releases. "Is this a problem?" he asked.

That's what friends are for, eh? Microsoft is unlikely to ever bring up criticism of Windows in its very own conference which is covered by journalists. Here is one report that references the above.

According to the report, Novell engineer James Bottomley referred to an internal study done by Intel which had found Linux performance had fallen by 2 percentage points at every release. Over the last 10 releases, the drop had been about 12 percent.

The numbers themselves come from Intel, but these were internal. Putting aside Intel's heinous crimes (yes, it commits crimes and then found guilty in multiple continents, by separate independent courts), there are many facts that Moblin fans prefer not see, such as Intel's support of SCO, relentless promotion of Vista 7, notoriously bad drivers for Linux, and collusion with Microsoft. See for example:

• Does the Intel/Microsoft Relationship Qualify as a Cartel?
• Intel-Microsoft Collusion in Sub-notebooks: A Smoking Gun
• Collusion Against GNU/Linux and Sub-notebooks: Part II
• Microsoft on Intel's Anti-Linux: “Please Keep Confidential. This is a Nightmare”
• Microsoft Accused of Profiteering ($1.5 Billion) from Crime with Intel
• A Gradual Fall of W|Intel... Thanks to OLPC?
• A Federal Court Judge Might Force Microsoft to Pay Up Money It Hardly Has
• Microsoft Accused of Profiteering ($1.5 Billion) from Crime with Intel
• William Poole of NComputing (Formerly Microsoft) Blamed in the Big Collusion
• Might Steve Ballmer be Deposed for Collusion After All?
• The Return of Vista Collusions: Will Ballmer be Deposed Soon?
• Eye on Microsoft: Intel Collaborations, Dumping, and Search Redux
• The Collusion/Collaboration Club Strikes Again
• SpikeSource is Not Open Source
• It's Not Innovation, It's Collusion
• Do Intel and Microsoft Use SpikeSource to Fight GNU/Linux?
• What We Can Learn About Novell from Intel-Microsoft-Dell-Hewlett-Packard Collusions

Intel would love to publicly appear like a friend of everyone (including Apple, which is a major client now). But Intel is still in the business of selling the most chipsets and there is no x86 supporter like Windows, which is almost exclusively about x86. Intel found itself asleep on the wheel with OLPC and ARM gaining traction, so it is currently trying to tie Linux to x86 by all means available, even Moblin.

As for the claims themselves, knowing that benchmarks are prone to self-serving fraud (Microsoft does this all the time [1, 2]), we have looked at Phoronix benchmarks of Linux 2.6.30, Linux 2.6.28, Linux 2.6.26, and Linux 2.6.23. Linux does not appear to be getting slower. We asked Michael Larabel if it is true that Linux gets "bloated" and he says that "it depends upon the area. Some areas of the kernel have slowed down while others have improved."

"Michael Larabel was right," claims our regular reader Oiaohm, who is somewhat of a Linux expert.

“Let us remember that Novell's Greg K-H self-servingly chose a criterion by which to slam Canonical and poison people's minds against Ubuntu.”So what is it that Novell's Bottomley was referring to specifically? Boot time has definitely improved, the file systems get faster as well, but it is always possible to find some specific test/s to suit whatever hypothesis is carved in stone and then requires proving. Let us remember that Novell's Greg K-H self-servingly chose a criterion by which to slam Canonical and poison people's minds against Ubuntu. Canonical contributes a lot outside kernel space, so Greg's smear ended up looking dishonest. Compare this to Greg's sweet talking when Microsoft's contribution to Linux was just a driver for Microsoft products [1, 2, 3, 4, 5, 6, 7, 8, 9]. Microsoft is his employer's ally, unlike Canonical.

In conclusion, adds Oiaohm: "Linux kernel is not just black and white numbers. Overall measurements are showing slowing [but] altering particular settings and setting up systems different ways cause completely different performance."

Referring to another regular reader, Oiaohm informally writes that this is the "reason why Diablo-D3 and me went head to head recently. I have been using cgroups to manage my processes so avoiding lot of cfs hell. Causing by auto grouping into users. So I am seeing way different bench numbers to what Diablo-D3 is getting, even when we are using the same source. It's also the numbers of merges over that time [...] Most of the independent trees to the main Linux kernel are no more [and the] Price of unifying it takes quite a few versions for all the side effects to be found and corrected. At some point something better in driver detection for hal searches has to be found. The merges into the Linux kernel should slow down soon [when] You get a linux feature list it is getting fairly complete." ⬆

"Microsoft did sponsor the benchmark testing and the NT server was better tuned than the Linux one. Having said that, I must say that I still trust the Windows NT server would have outperformed the Linux one."

--Windows platform manager, Microsoft South Africa
Reference: Outrage at Microsoft’s independent, yet sponsored NT 4.0/Linux research