LWN: Comments on "RHEL clones: a surfeit of riches"

RHEL clones: a surfeit of riches

sudhirgandotra — Thu, 08 Mar 2012 11:27:25 +0000

Could you finally create such a script to automate the re-spinnig of a distro like RedHat/CentOS ?
This can help a lot of students to learn the internals of Linux and spread it further.
I would be intereste in such a script of tutorial.
thanks.

RHEL clones: a surfeit of riches

filteredperception — Wed, 11 Jan 2012 20:56:04 +0000

Actually you did see the case I was really focused on, i.e. modified binaries generated from unmodified source. I.e. rebuilt on top of a differing linux distro (gcc/etc) platform. And in that case, Dave made it perfectly clear that redistributing such things, under any sort of moniker whatsoever, would be something he would prosecute, if for no other reason than the IP law angles that involve needing to prove that you have actually enforced your trademarks. Basically he was saying- if you want to redistribute, you'll have to scrub every protected mark, and for legal reasons he explicitly (like many other corporations) chose not to specify what exactly that means (trademark names in email addresses within source code files? exactly where is the line drawn and how is the novice university-level GPL enthusiast to know how they need to draw that line for legal/ethical reasons?). In the end I used some wicked regexs and even sillier methods to just go overkill scrubbing devkitpro marks. But it was more technical work than I would want to wish upon the average open source user, modifier, and redistributor. Now, on the other hand, I'll give RedHat credit for the sadly practical hands off method they seem to be taking on that account. I.e. not seeming to care about things like even the bugurl and whatnot found in e.g. the ScientificLinux Xorg srpm's specs build code. And in general Fedora over the years clearly seems to be heading in the right direction, and making excellent progress, in facilitating the ability for someone to rebuild from source, remove the needed amount of fedora marks, and redistribute without having to gain anyone's permission. (Though of course that isn't really true, since if you wanted to enhance/modify any code amongst the subset belonging to firefox, you'd have to go to extraordinary superhuman lengths to scrub the firefox marks for the same reasons. ... sigh ...

RHEL clones: a surfeit of riches

nybble41 — Wed, 11 Jan 2012 20:31:22 +0000

> users have no right to redistribute, even unmodified copies of GPL binaries including protected trademarks

In the case of unmodified copies, that would be a purely descriptive use of the trademarks, and thus normally permitted under trademark law. After all, you're using the trademark to designate the precise binaries distributed by the trademark holder, and protecting that association is the entire purpose of the trademark.

Distributing modified binaries would, of course, be an entirely separate issue, even if they were derived from the same source code. "Based on TrademarkedName(R)" would probably pass, as a purely factual statement, but you can't legally label your own version with someone else's trademark and pass it off as the original software.

RHEL clones: a surfeit of riches

filteredperception — Wed, 11 Jan 2012 19:57:21 +0000

"the GPL then gives you the general right to copy"

Hmm? But asside from my skepticism of what truth there actually is in that, you fail to mention the immensely broad realm of 'non-private, non-commercial' redistribution. That matters a lot to me. So much so that I spent an obsessive amount of time scrubbing trademarks from DevKitPro's redistribution of GCC as a Nintendo-DS homebrew development environment. Specifically, Dave Murphy educated me as to the fact that users have no right to redistribute, even unmodified copies of GPL binaries including protected trademarks. Basically one of my overriding goals is to provide a turnkey LiveOS distro that includes both the code of RHEL6 and DevKitPro, but in a form in which any user, can _easily_ fix a 1-line bug, and redistribute, without the need for permission from any party, and even presuming that I place trademark redistribution protections on my derived work equivalent to the way RedHat and DevKitPro do. That is, I believe, quite in tune with the spirt of the GPL. And you'd be amazed what a pain in the ass that task is to actually accomplish. But it'll happen soon enough...

RHEL clones: a surfeit of riches

paulj — Wed, 11 Jan 2012 07:23:27 +0000

The RHEL EULA certainly seems to place "RedHat Enterprise Linux" under the GPL:

http://www.redhat.com/licenses/rhel_rha_eula.html

It asserts RedHats' rights over its trademarks and the need to acquire permission in any commercial redistribution. It doesn't mention private redistribution, but the GPL then gives you the general right to copy, and private redistribution isn't restricted by trademark law.

RHEL clones: a surfeit of riches

filteredperception — Wed, 11 Jan 2012 01:33:12 +0000

"I think RedHat even explicitly licence the OS in the aggregate under the GPL"

I doubt this is the case, probably for several reasons, not the least of which would be because the spirit of the GPL, and perhaps even the current interpretation of it would suggest then that RedHat be required to produce the build scripts used to generate that aggregation, so that others could easily modify and then redistribute an enhanced/changed version of that aggregation.

One of the subtler angles I see of all this, is how we are, IMHO, probably well past the point of not thinking of linux distros as singular derived works in their entirety, rather than as collections of objects that can be thought of as derived works, absent the collection being thought of as such.

That said, RH does go a long way towards empowering users to create derived from-source distributions. It's just that 90/10 thing and how that last 10% of the work just gets fractally gnarly. Probably I just made the mistake of thinking koji was the right tool for the job I was using it for. I would like to leverage it to enable clustered superfast builds, but given I think my acer aspire one netbook(amd dual core, 4Gram) can probably build all of current rhel6+updates from source in a week... eh...

RHEL clones: a surfeit of riches

paulj — Fri, 06 Jan 2012 23:49:48 +0000

You're perfectly entitled to run RHEL on your home systems, if you can get someone with RHEL CDs to make some copies of them (which they are also perfectly entitled to do for the RHEL CDs - I think RedHat even explicitly licence the OS in the aggregate under the GPL). Of course, the home RHEL user won't have access to any updates whatsoever.

RHEL clones: a surfeit of riches

drag — Fri, 06 Jan 2012 15:18:40 +0000

> What I have never understood is why RH does not provide a personal/non-profit way for someone to run RHEL.

They do. It's called ftp.redhat.com. People have taken advantage of it and that is why we have things like CentOS and Scientific Linux.

RHEL clones: a surfeit of riches

anselm — Fri, 06 Jan 2012 12:58:38 +0000

So Red Hat decided to switch to a model where pay-in-advance is the only way to get the distro.

Actually I've heard that Red Hat will be happy to support your existing CentOS installation if you buy the requisite RHEL licence(s), with no need to reinstall everything.

RHEL clones: a surfeit of riches

walex — Fri, 06 Jan 2012 12:30:22 +0000

But Red Hat are a for profit company, therefore as a company they decided to focus on their paying customers for EL. The non-profit variant is Fedora and it is deliberately not EL-like.

Part of the problem is that as another commenter points out a lot of very clever sysadms liked to play the game of looking good to management by using a business distro without the cost of support, reckoning it would be infrequently needed, only to turn around and blame the business distro when support became needed. So Red Hat decided to switch to a model where pay-in-advance is the only way to get the distro.

RHEL clones: a surfeit of riches

walex — Fri, 06 Jan 2012 12:21:24 +0000

I think a fully automated way to build a base distro from sources as you describe would be very welcome at Scientific Linux.

Their main mission is to support the CERN WLCG Grid and they are somewhat overworked. They had therefore considered adopting CentOS as the base and then work only on the "Scientific" bits.

Try to submit a presentation at HEPiX, it might be well received. The CERN guys might be especially interested, as they build their own slightly customised version on top of Scientific Linux, and are the guys who really matter.

RHEL clones: a surfeit of riches

michel — Fri, 06 Jan 2012 11:50:35 +0000

Well, the problem there is that people bought the box, which implies at least some sort of support and is not unreasonable for the people who purchased the product to expect. And I think there are ways to deal with support calls from folks that have no support, but I guess this is why I am not running a $1B company.

RHEL clones: a surfeit of riches

ewan — Fri, 06 Jan 2012 03:36:05 +0000

What you're suggesting is essentially the situation that existed with the old Red Hat Linux distribution - people could buy box sets or download ISOs, but support was extra. As I understand it, the problem was that a lot of people were contacting Red Hat for support, despite not having support contract, simply because they were running 'Red Hat Linux'. The load and the bad feeling caused by turning people away wasn't doing Red Hat much good, so they moved to the current RHEL model where you don't even get a copy without having a support contract. The rules for the rebuilds are that they have to remove all the Red Hat branding, so that anyone faced with a copy of CentOS, SL, etc. has no reason to get the impression that they're running something that has anything to do with Red Hat at all.

RHEL clones: a surfeit of riches

filteredperception — Thu, 05 Jan 2012 18:38:59 +0000

"That list just reinforces my impression that there are only two free RHEL rebuilds that matter: CentOS and Scientific Linux. Others should be ignored."

The reason I chose to work on Ascendos, instead of being content with with 'the big two', was because they weren't meeting, or interested in meeting my needs/wants in an *el clone/fork. Specifically, I'm a paranoid sort that doesn't believe an 'enterprise linux' build has any business being done on an online system, or in a way that couldn't be reproduced by a military grunt from a simple recipe/instructions/documentations/script.

CentOS is very clear about not wanting to enable that, because were that possible, people could trivially fork/rebrand/rebuild. And then the ego stamped into the name/logos/theme of their rebuild would not be enough to keep them relevent as an organization (or so they appear to feel).

ScientificLinux on the other hand has the problem in my mind of being the output of an inexusably* unreproduable result of a manual[1] build process. Hopefully done on offline systems, but I wouldn't be surprised if that is not the case.

Thus I spent a lot of time trying to develop an 'el-build' script which
would simply

a) download upstream sources and minimal required bootstrapping environment
from mirrors.kernel.org/archive.fedoraproject.org.

b) apply a few megabytes of deltas/patches to transform the collection of
srpms into a newly forked/rebranded set

c) build into a mirrorable distro.

The key being to encapsulate the entirety of the differences from upstream
into the smallest possible set of deltas and build scripts, such that they
are as easy to audit and rebuild as humanly possible.

That was my vision, FWIW. But yes, I did fail to attract other developers,
though the only ones I really saw in the area seemed too much in my mind to
just want to become the next centos-like technocrat clique, instead of my
goals of just automating away the whole [expletives deleted] problem so that even amateurs could remove redhat's/centos's ego-stamp from their distro, and replace it with their own.

-dmc

* from the perspective of traditional 'scientific method'

[1] http://listserv.fnal.gov/scripts/wa.exe?A2=ind1111&L=...

RHEL clones: a surfeit of riches

michel — Thu, 05 Jan 2012 14:20:15 +0000

What I have never understood is why RH does not provide a personal/non-profit way for someone to run RHEL. Simply allow you to run it, without any support, but with access to updates, etc. Clearly, I'm not going to pay for enterprise support for a bunch of small home based systems.

I use Fedora for my main machine at home, but end up using SL for some of my servers to get the more slowly evolving/stable aspects that I want there.

Perhaps it's simply a scaling problem.

RHEL clones: a surfeit of riches

eru — Thu, 05 Jan 2012 06:16:15 +0000

That list just reinforces my impression that there are only two free RHEL rebuilds that matter: CentOS and Scientific Linux. Others should be ignored. These two also seem to have sufficiently different niches to justify both of them. CentOS (which I run on several machines) tries to be maximally RHEL-compatible, and Scientific Linux feels free to add interesting stuff, while keeping the RHEL base for stability.