Slashdot Mirror
IBM Refuses To Certify Oracle Linux
Andrew writes "IBM has thrown a spanner in the Oracle Linux works by refusing to certify that IBM's software portfolio will run and be supported on Oracle Unbreakable Linux. If IBM applications turn out to be incompatible with Oracle Linux, then it will be up to Oracle to resolve any issues. This conservative stance of IBM's is unlikely to help Oracle sell Linux subscriptions to businesses that use any of IBM's large software portfolio."
Now that Oracle has added "Unbreakable" in front of the word "Linux"... Linux is finally going to become insecure :(
Note to Linux developers: remember to add all your SVN commits as cron jobs, and forward date them all 2 years, or 3 years if they're critical security patches.
IMNSHO, db2 pisses on Oracle from a great height.
Sent from my ASR33 using ASCII
I'm fairly sure that if someone offers enough money, they could have that assurance on Oracle, Ubuntu, or anyone else's Linux too.
I don't recall any of Rick Stallman's lectures being about charity.
Oracle is trying to shift the blame from their software stack to IBM's before they've even deployed a box. In other words, if you have a problem and are running IBM software, Oracle wants IBM to foot the bill of researching the bug or issue.
Who is Oracle to dictate that problems are automatically some other vendor's fault instead of their own?
I do not fail; I succeed at finding out what does not work.
but big enterprise doesn't think that way, they want certified compatibility and adherence to the letter of support contracts. Good luck calling EMC or Hitachi and saying your CentOS has problems accessing your 20TB disk array with a given HBA and switch. You'd void your warranty right then and there. As an aside, CentOS lags RedHat in patches, and also has to rewrite parts of the redhat admin system, it isn't 100% the same.
This is hardly unusual. Companies spend a lot of money to certify software distros, and put their own maintenance dollars on the line when they certify them.
A third party saying "use our stuff--it's just the same as theirs" isn't necessarily credible. Maybe they're a clone, or maybe they're a clone today and might not be tomorrow. Or maybe they'd only clone part of the distro, leaving out critical parts. Or maybe they'll add custom stuff to the distro. IBM isn't under any obligation to believe Oracle's marketing materials and automatically certify based on taking Oracle's word that "it's the same and always will be."
There are DOZENS of RHEL clones out there (CentOS is the most popular, but hardly the only one). I don't think IBM considers any of them "certified."
Actual quote from TFA: "We are going to wait and see if there is traction in the marketplace," McMahon said. "If clients want it (Oracle), then we will support it."
This is a non-issue, and someone's using the "IBM vs. Oracle!" angle to generate traffic and controversy by stirring people up. Looks like they succeeded.
No offense, but as a hardware vendor I'd do this too. Because otherwise, where do you stop? "Yeah, we're running on this custom-compiled Gentoo with a few third party extensions. We're seeing some errors in our custom logs that look like our proprietary apps can't connect to your hardware. Send an engineer."
As a vendor, I will tell you "OK, we've checked out and certified that we work with these distros. Anything else, it will probably work but you're on your own if it doesn't." Seems reasonable to me.
Now, if you're concerned that vendors will use this to shut out "free" distros from being supported, maybe that's a case worth making. But some of this is market demand--if HP kept getting the question about "hey, will you support this on CentOS?" from hardware customers, and were losing customers by saying no, you can be they'd look into CentOS support. They are not vested in propping up RedHat's licensing business.
IMO, the main issue that big companies have and will continue to have with Linux is distro fragmentation. It's just not feasible anymore to test your applicaiton/server/hardware with every conceivable distro that's out there. So you pick some, and those are the ones your customers tell you they're running.
Put another way, *I* could come out with a RHEL clone distro tomorrow. Are you telling me it would be reasonable for me to expect HP to support it?
"Good luck calling EMC or Hitachi and saying your CentOS has problems accessing your 20TB disk array with a given HBA and switch."
I do this all the time, I run CentOS in development and most of the test environment. When I see a problem on CentOS, I verify that the problem exists on one of my RHEL test boxes, and call them up. When they fix the problem on RHEL, it is either automatically fixed on CentOS, or I replicate whatever they did on RHEL on CentOS and the problem is fixed. You just have to learn how to play their game, if they say they only support "expensive X", then have as few of "expensive X" around to satisfy that requirement.
"it isn't 100% the same"
It's enough the same that I have never run into anything that broke on one that didn't break exactly the same way on the other. CentOS is so good that I have started to move some of my production systems to it, but I will always keep a fair number of RHEL boxes around, since third parties need someone to point a finger at when they determine that it isn't their stuff.
I just hate the whole concept of Linux distribution certification, because it tells me that there's something wrong with running software on it.
For most IT dept.-written apps that rely on super-common well-known library functions, no, the distribution doesn't make much of a difference. But once you start doing lower-level stuff (like the sort of stuff every software application IBM sells does), things start to not work right.
This is a big problem with Linux, and no amount of wishful thinking will make the problem go away. Apps not working on all distributions is exactly the sort of problem that the Linux Standards Base (went nowhere) and United Linux (supported by Caldera/SCO) were supposed to prevent.
Organisations generally just try it out on a newer version of an OS, and if it works OK in a trial period (even if they have to tweak things to get it to work) they go with it, and they don't fly into a massive panic.
Maybe some IT shop that doesn't care about their software actually working can do that, but actual software companies that make their living selling software MUST perform testing.
Yes, there are many organizations that do that, but those are either small and/or low-quality IT shops and/or non-critical apps.
I have several healthcare industry customers that are running OS software that is coming up on three years of ageing out of OS vendor support because their app vendor STILL hasn't certified a more recent O/S version. For them, and most customers, the app vendor support is far more important than OS vendor support, because they know that most day-to-day bugs are in their apps, not their OS. Personally, I know that I crash Mozilla (and other apps) a heck of a lot more than I have ever crashed Windows.
If Oracle can say "Yes, this will run" to their customers, and their customers try it out and it does actually run, then no one will care.
Those customers will care very much when they try and call IBM to receive assistance under their support contract for their expensive and complex application and IBM says "Sorry Mr. Customer, you are running in an extremely unsupported and untested environment." Usually this will be accompanied by some limited best-effort support to make sure that it is not an obvious bug in the product.
Now if enough customers ask for it (and are willing to pay), I am sure that IBM will be more than happy to certify their apps on Oracle Linux. Yes, Oracle is a competitor, but so is M$, and plenty of IBM software runs on Windows. But IBM is not going to go out and certify Oracle Linux just because Oracle is whining about it. I am equally sure that if IBM rolled out their own distro tomorrow, Oracle would not be falling over themselves to certify their apps for it either.
This whole concept of distributors and software vendors protecting themselves (and engineering some lock-in, incidentally) by certifying, or certifying for, certain distributions just isn't helping Linux or open source software get more widely used.
The fact of the matter is that there ARE differences between distributions, and those differences have been known to break a lot of applications. Because of this, there is no way for a software vendor to get around distribution certification. If you certified your mega-dollar application to run on any Linux distro, what do you do the first time some clown calls up with some home-grown hybrid of five different distros and wonders why it doesn't work?
Software companies are in the business of making money, not "helping Linux or open source software get more widely used." If Linux distro writers want to make the burden of application certification easier, then the onus is on the Linux folks to get their act together and make Linux distros more homogenous. Don't blame the software vendors for this sorry state of affairs.
SirWired
Oracle isn't selling to IT people; they're selling to IT peoples' managers.
- "History shows again and again how nature points out the folly of men" -- Blue Oyster Cult, 'Godzilla'