Oracle recently posted “Oracle Linux: A better alternative to CentOS,” which is a really interesting page – it’s well done, I suppose, in and of itself, but some of the details are rather sketchy and there’s a giant red herring that doesn’t seem to be acknowledged.
Here’s the red herring right here:
…it means virtually no delay between when Red Hat releases a kernel and when Oracle Linux does…
So… wait a second. Oracle is making an argument based on the concept that they – Oracle – release a new Linux kernel as quickly as Red Hat does, while CentOS is less likely to do so. They’re using a graph with data from 2011 to show this:
Interesting. So the last datum used showed CentOS being faster than Oracle – which means their point can’t be the most recent data point (which is wise, honestly) but is instead about variance.
“See,” they seem to say, “in 2011 CentOS was slower, with a wide variance. They consistently got better, but… the variance! Look at the variance!”
The problem here is that this highlights an issue with CentOS’ organization – and from the looks of it, things were getting better. The variance might be significant, but it also might mean they had problems in the beginning, but straightened them out.
But there’s a bigger deal here, larger than “CentOS or OEL?”
Why bother with either one? Oracle seems to be saying “Use Oracle Enterprise Linux” because it’s closer to a desirable norm — which isn’t Oracle or CentOS. It’s Red Hat Enterprise Linux itself.
It’s like this: you have two detergents, let’s call them “Smash-O” and “UltraKleen.” UltraKleen says “We’re better than Smash-O because we’re not fluorescent, and we’re almost as good as DirtAway. When DirtAway improves, so do we, to catch up!”
What this statement is really telling you is that if you’re looking for a great detergent (in scope, of course), you are best off going direct with DirtAway — it’s setting the standard, not Smash-O or UltraKleen. Both of those are playing “catch-up” with DirtAway.
[aside]Remember: 98.6152% of all statistics are made up on the spot.[/aside]
They’re making an appeal based on a different variable than quality, such as price point. Perhaps they cost 75% of what DirtAway does, therefore you’re willing to accept that your clothes are perhaps 80% as clean.
Oracle is doing exactly the same thing: they’re suggesting that CentOS is the “Red Hat Enterprise Linux”-Lite, and that if you’re looking for the “Lite” version of an Enterprise Linux, that you might as well be using Oracle Linux, because kernel updates have a smaller standard deviation times in the release cycles.
You’re still depending on Red Hat for those release cycles.
But if you’re settling for a “Lite” version, then Oracle’s okay for that!
In other words, they’re giving Red Hat Enterprise Linux (and Red Hat) the highest form of flattery: imitation.
In case you’re interested in another comparison, here’s the Oracle graph, with an extra set of data points: I added RHEL’s kernel releases to the data. Note: the standard deviation is zero for every data point.
So if a consistent, low deviation for releases is your desired result… 
{ 1 comment… read it below or add one }
Nice write up and the point well made!