This post appeared on another forum:
Will CentOS become obsolete now because of the changes Red Hat is implementing?
Red Hat has changed the way it distributes Enterprise Linux kernel code in an effort to prevent Oracle and Novell from stealing its customers, making it more difficult for these competitors to understand which patches have been applied where.
Some have speculated that the change is designed to make it harder for Oracle as well as the open source CentOS project to build their own Linux distributions. But Stevens says this is not the case. He says the change is meant to hamper Oracle and Novell's ability to offer support to customers who are already running Red Hat Enterprise Linux (RHEL).
"We made the change, quite honestly, because we are absolutely making a set of steps that make it more difficult for competitors that wish to provide support services on top of Red Hat Enterprise Linux," Red Hat chief technology officer Brian Stevens tells The Register, before naming those competitors. "Today, there are two competitors that I'm aware of that go to our customers directly, offering to support RHEL directly for them...Oracle and Novell."
In essence, Red Hat is trying to hide information from these competitors that is essential to providing support for RHEL specifically. "What we're trying to impede is competitors that come to customers who are already running RHEL under subscription from Red Hat and saying 'Don't pay Red Hat anymore, pay us, and don't make any changes to your systems'," Stevens says.
He insists that the change does not violate either the letter or the spirit of RHEL's GPL open source license. "We were very careful that what we've done does not impede what our customers need to accomplish or what the community needs to accomplish." And he says that the change would not really hamper the development of other Linux distros, including CentOS.
"We haven't at all restricted CentOS's ability to grab source code and recompile it and clean-out trademarks and package it. It's just some of the knowledge of the insides that we're hiding," he explains. One longtime CentOS developer agrees.
"I'll not lose sleep on the matter," CentOS co-founder Russ Herold tells The Reg.
In November, with the release of Red Hat Enterprise Linux 6, the company released its kernel package with all patches pre-applied. "In the past, we distributed the kernel as a base file and then a set of add-on patches that accompany it. Then when you did a build, the build process automatically applied all those patches to the kernel file," Stevens says. "Now, we integrate those patch files directly into that kernel. We do the first part of the build process prior to distribution."
This was recently noticed by kernel-community member and LWN editor Jonathan Corbet, who took issue with the change, calling Red Hat's package "obfuscated" kernel source code.
"Distribution in this form should satisfy the GPL, but it makes life hard for anybody else wanting to see what has been done with this kernel," Corbet wrote. "Hopefully it is simply a mistake which will be corrected soon." Others speculated that the move would undermine not only Oracle's Unbreakable Linux, but also CentOS. Both are based on RHEL.
CentOS is meant to be a RHEL clone. Whereas the compiled bits of Red Hat Enterprise Linux are only available under a Red Hat paid subscription, CentOS is completely free.
"The changes will make work harder for distributions such as CentOS, the community-built Linux distribution ... based on Red Hat's sources," H Online said. "CentOS is built from the RHEL source by a limited number of volunteers and Red Hat's change in policy will mean more work for them unless more volunteers or other companies step in and provide them with assistance."
We heard similar noises from an experienced Linux kernel developer. He said that Red Hat's change was like shuffling all the cards in an old fashioned Dewey Decimal library file system – the card you want is still there, but finding it is no easy task – and that this would cause problems for CentOS, which is an economic threat to Red Hat.
But CentOS founder Russ Herold insists the change is not a big issue. "Private local trial builds of the released RHEL 6 sources by me and others have proceeded with no major problems. I just do not see that the changes as some earth-shattering change. I just think [the patches will be] incrementally more difficult to figure out," he says.
"Nothing in Red Hat's new approach prevents a person from running a local version-control system, containing the pristine kernel at point A, and the Red Hat variant which we might call point B. Then one runs a 'diff' in that version-control system between A and B, and starts reading the diffs to see what is happening. Over time, both the pristine kernel, and the patched Red Hat versions will vary, and one will get a sense for which 'diff' parts matter, and which are cosmetic cleanups."
Other distros will not be affected, Red Hat's Stevens says, because the company distributes its kernel changes upstream as well. "The work that we've done should not impede companies from building their own versions of Linux and supporting those for their customers," he says. "All the code we deliver through RHEL is out there. In most cases, the changes that go into RHEL. We already distribute into the upstream kernel. We have an upstream-first policy, where we're developing openly and then later integrating into our tree and then delivering it. So it shouldn't at all impede the community or anybody that's in the business of competing on that."
Red Hat, he reiterates, is trying to keep RHEL-specific knowledge away from Oracle and Novell. With past RHEL kernel-code distributions, the patches mapped to articles in Red Hat's knowledge base. "It makes competitors do heavy lifting," he says. "If you want to support RHEL, remove the trademarks, and do some heavy lifting. If nothing else, it causes competitors to have to invest."
This won't hamper CentOS, he says, because CentOS isn't in the support business. "The code is still available. It's just more difficult to support the distro as a commercial entity. CentOS is not in the support business."
Oracle and Novell are in the support business. And whatever collateral damage was caused by Red Hat's change in policy, one thing is for sure. On some level, it will indeed be more difficult for Oracle and Novell to pilfer the company's customers.
Full story here : http://www.channelregister.co.uk/201...ode_packaging/
Can any of the CentOS team please comment on this?
P.S. I don't want to put extra pressure on you guys seeing as the recent conversations about Centos 5.6 & CentOS 6 have done enough damage already, but it would be interesting to know what the future of CentOS is like with RedHat's change
On 05/03/11 10:07, Rudi Ahlers wrote:
This post appeared on another forum:
Will CentOS become obsolete now because of the changes Red Hat is implementing?
<snip>
But CentOS founder Russ Herold insists the change is not a big issue. "Private local trial builds of the released RHEL 6 sources by me and others have proceeded with no major problems. I just do not see that the changes as some earth-shattering change. I just think [the patches will be] incrementally more difficult to figure out," he says.
"Nothing in Red Hat's new approach prevents a person from running a local version-control system, containing the pristine kernel at point A, and the Red Hat variant which we might call point B. Then one runs a 'diff' in that version-control system between A and B, and starts reading the diffs to see what is happening. Over time, both the pristine kernel, and the patched Red Hat versions will vary, and one will get a sense for which 'diff' parts matter, and which are cosmetic cleanups."
<snip>
Full story here : http://www.channelregister.co.uk/201...ode_packaging/
full non-truncated link:
http://www.channelregister.co.uk/2011/03/04/red_hat_twarts_oracle_and_novell...
Can any of the CentOS team please comment on this?
Which part of Russ Herold's ("CentOS founder") comments above did you not read?
Come on, this whole story is total nonsense and has been responded to a number of times. Red Hat are legitimately protecting their business model against competitors (namely, Oracle and Novel) and the changes have no impact towards rebuilders. This is a good thing - if Red Hat doesn't exist the CentOS doesn't exist.
On 03/05/2011 04:07 AM, Rudi Ahlers wrote:
This post appeared on another forum:
This kernel change does not impact the ability to rebuild the source as is, it just makes it much harder to do anything except build the pristine kernel from kernel.org or the Red Hat kernel.
You can still compare the RH tarball to the pristine kernel and see all the changes ... they are just not added one at a time.
This will make it very difficult to back out one patch (because your server is running version xyz of farmer joe's ethernet card, and that does not work with PatchA).
This should not impact building the kernel ... it might impact things like the CentOSPlus Kernel or CentOS providing a "stop gap" kernel (in the testing repo) while waiting for Red Hat to correct a problem and get their kernel through engineering and then released.
That is not to say I like the changes, as it will have impacts ... but as long as they only do it to the kernel, it is not a big problem.
If they do it to every package in the OS, then that would be a much bigger problem. It is much easier for us to find and remove trademarks when they are inserted via a patch into existing upstream code ... if we only had the modified tarballs and not the patches, it would be much harder to find all the things that need changing.
On 03/05/2011 04:22 AM, Johnny Hughes wrote:
On 03/05/2011 04:07 AM, Rudi Ahlers wrote:
This post appeared on another forum:
This kernel change does not impact the ability to rebuild the source as is, it just makes it much harder to do anything except build the pristine kernel from kernel.org or the Red Hat kernel.
You can still compare the RH tarball to the pristine kernel and see all the changes ... they are just not added one at a time.
This will make it very difficult to back out one patch (because your server is running version xyz of farmer joe's ethernet card, and that does not work with PatchA).
This should not impact building the kernel ... it might impact things like the CentOSPlus Kernel or CentOS providing a "stop gap" kernel (in the testing repo) while waiting for Red Hat to correct a problem and get their kernel through engineering and then released.
That is not to say I like the changes, as it will have impacts ... but as long as they only do it to the kernel, it is not a big problem.
If they do it to every package in the OS, then that would be a much bigger problem. It is much easier for us to find and remove trademarks when they are inserted via a patch into existing upstream code ... if we only had the modified tarballs and not the patches, it would be much harder to find all the things that need changing.
It's my sense that Redhat views CentOS and SL as benificial for them and not as competition. I think over time many people install CentOS and it leads to them selling more redhat licenses from companies who really do need support. Our testing and filing of bugs in their database is also supportive of the development process in the same way that it is with fedora.
Over the years, I've talked several times to redhat sales people and they've always recognized people that contribute to the Open Source community and were not really in need of a support contract and told them to just download Open Source distributions. I've never once felt pressure from them to replace my CentOS systems with Redhat licenses, though I've purchased a few over time.
I think redhat is a great example of an open source company that truly gives back to the community without fear of losing business as a result. Their concerns around these two large commercial competitors are legit and it is apropriate for them to be addressed. We all benefit from having companies like Redhat around.
Nataraj
On Sat, Mar 5, 2011 at 5:07 AM, Rudi Ahlers Rudi@softdux.com wrote:
This post appeared on another forum:
Will CentOS become obsolete now because of the changes Red Hat is implementing?
Red Hat has changed the way it distributes Enterprise Linux kernel code in an effort to prevent Oracle and Novell from stealing its customers, making it more difficult for these competitors to understand which patches have been applied where.
Oh, son of a ...... . I once spent six months *un-obfuscating* a lot of kernel code changes in a company's internally used and optimized kernel, becuase the developers had been doing their developments off in la-la-land. It's a good story, ask if you're curious.
Such tweaking makes it very hard to integrate other published kernel changes from local kernels or the upstream kernel. This is, of course, the *point*, but if you're someone like me who's sometimes asked to incorporate optimizations or drivers on top of someone else's code changes, it makes it considerably harder to tell which changes you might be overwriting or conflicting with.
It really shouldn't affect CentOS releases: the kernel code doesn't have lots of trademarks in it, so it shouldn't be hard to recompile the kernel without tweaks. But for someone like me who gets asked to support Infiniband (before it was popular) or various fiber channel cards with RHEL 4 or CentOS 4 for software compatibility reasons, it could be an issue.
yeah but he wants to hear from the horse's mouth....
Who knows what the guy really said when a reporter quotes something....
On Sat, Mar 5, 2011 at 9:22 AM, Lamar Owen lowen@pari.edu wrote:
On Saturday, March 05, 2011 05:07:10 am Rudi Ahlers wrote:
Can any of the CentOS team please comment on this?
One did, and he's quoted in the article referenced. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Sat, 5 Mar 2011, robert mena wrote:
yeah but he wants to hear from the horse's mouth....
Who knows what the guy really said when a reporter quotes something....
The 'pull quotes' attributed to me (the horse in question) by the Reg are accurate and carry good context
Anyone who follows this mailing list and -devel or my blog, has seen the balance of what I said, over and over. People who want to understand how to solve the issues of rebuilding ... need to roll up their sleeves and get down and dirty doing rebuilds, so they see and learn how to solve the issues. Talking about it, speculating, conducting 'thought experiments' about what _should_ happen do little that is useful
There is no magic bullet or shortcut to becoming a rebuilder, no 'secret sauce' as hughesjr says. Just systematic problem solving
Exactly ONE PERSON has followed through since I offered private guidance over three months ago, and that person now has a server oriented, trimmed down RHEL 6 sources rebuild. I offered pointers, but they learned by doing. I am proud to see the install CD sized ISOs for i386 and x86_64 rather than the bloat DVD images of RHEL (some surgery to anaconda was required, and that spin does not seek binary identical replication of the upstream, so it is not a CentOS clone or competitor. I'll probably offer the image at pmman soon)
It's not magic folks
-- Russ herrold
On Sat, Mar 5, 2011 at 4:57 PM, robert mena robert.mena@gmail.com wrote:
yeah but he wants to hear from the horse's mouth.... Who knows what the guy really said when a reporter quotes something....
On Sat, Mar 5, 2011 at 9:22 AM, Lamar Owen lowen@pari.edu wrote:
On Saturday, March 05, 2011 05:07:10 am Rudi Ahlers wrote:
Can any of the CentOS team please comment on this?
One did, and he's quoted in the article referenced. _______________________________________________
Exactly, I don't know how "twisted" that article really is. And it would be good to hear from our devs directly :)
On Mar 5, 2011, at 11:25 AM, Rudi Ahlers wrote:
On Sat, Mar 5, 2011 at 4:57 PM, robert mena robert.mena@gmail.com wrote:
yeah but he wants to hear from the horse's mouth.... Who knows what the guy really said when a reporter quotes something....
On Sat, Mar 5, 2011 at 9:22 AM, Lamar Owen lowen@pari.edu wrote:
On Saturday, March 05, 2011 05:07:10 am Rudi Ahlers wrote:
Can any of the CentOS team please comment on this?
One did, and he's quoted in the article referenced. _______________________________________________
Exactly, I don't know how "twisted" that article really is. And it would be good to hear from our devs directly :)
There's been some traction on the SL list about this as well, pretty much saying what was quoted from Centos.
Basically no big deal other then perhaps having some impact centosplus kernel.
The article seemed pretty clear to me but perhaps I missed something?
- aurf