Specifically, Mark fails to mention that Centos is listed separately from RHEL in the survey, and when added together they come to 41%, more than twice that of Ubuntu.
I'm a Canonical fan, I'm typing this on 12.04 right now, but that seems a bit misleading.
It's also misleading because he talks about "large scale enterprise workloads," when he really should mention that their marketshare is almost entirely Linux desktop workstations. Most server workloads are entirely RHEL and CentOS.
CentOS and RHEL combined total 41.1% of all websites on the Internet. Ubuntu is a distant 3rd or 4th place behind Debian, depending on whether you count RHEL and CentOS as the same OS, or separate.
Mark's response is kind of bogus: it's true that Fedora shouldn't be lumped with RHEL unless you lump Debian with Ubuntu, but you can't say this for CentOS. It's literally RHEL with all the branding removed (and no support from RedHat).
However, the graph does point to a decline in the willingness to pay lots of money for a server Linux distro, which is an interesting trend.
If it's maketshare I assume the graph is in % and it could mean that there is a lot more servers being added and not a decline in installs of RHEL. I can see that with Amazon EC2 instances and other cloud platform Ubuntu would be really popular on those. Buying a RedHat license for you VPS instances might not be worth it but then CentOS vs Ubuntu would be more interesting there.
Is it just a question of paying lots of money for a distro? Or is it a question of the market decentralizing, and so vendor-centric solutions doing increasingly poorly?
BTW, this back and forth proves what Heisenberg forcefully argued in his book, "Physics and Philosophy," that theorizing fundamentally involves projecting your own assumptions onto data you see. We all see the same graph. We get to argue to no end as to what it means.
Debian and ubuntu are clearly two substantially different distributions with different packages, policies and goals. As are RHEL and Fedora. The only difference between Centos and RHEL is some branding and whether you paid a license fee.
Those same, clueless organizations prohibit use of Ubuntu, too, because RHEL is the only thing they've certified. None of this changes how badly misleading and disingenuous Mark's statements are.
On the other hand, the Debian ecosystem (of which Ubuntu is primary beneficiary, even though not the primary contributor) is probably about the same size.
There's one other important difference: many enterprise applications are certified for RHEL but not CentOS, therefore IT departments in many large organizations are not allowed by internal policy to use the latter for anything critical.
(FWIW, Ubuntu is always the same exact OS regardless of whether the customer hires Canonical for support or not.)
We would love to sell and use Ubuntu based systems to the military and some enterprise but unfortunately there is a slew of certificates that RHEL systems have that Ubuntu don't. Like FIPS 140-2 etc.
Is Ubuntu a supported platform for proprietary enterprise software & hardware the way RHEL is? Most of the time i've seen RHEL picked not only for its own support contracts but because it's supported by vendors like Oracle, Dell, RSA, VMware, etc. I'm sure apt has the same core features as rpm/yum, and probably similar stability with regards to patchsets, but vendor support is the killer enterprise feature for any distro.
(P.S. it's true that CentOS is used a lot in place of RHEL for cost cutting; they just change the redhat-release files and RPM provides and lie to the vendors about the platform)
Specifically, Mark fails to mention that Centos is listed separately from RHEL in the survey, and when added together they come to 41%, more than twice that of Ubuntu.
I'm a Canonical fan, I'm typing this on 12.04 right now, but that seems a bit misleading.