Weekly Edition
Return to the Front page
| |||||||||||||
An Evening with Bruce Perens
Bruce Perens was in Denver this week for IBM's Linux strategy briefing and offered to speak to the Colorado Linux Users and Enthusiasts (CLUE) Linux User Group the night before the IBM event.
The talk was billed as "The Future of GNU/Linux and Free Software," but
Perens talked a great deal about the history of free software as well.
After covering his history with Linux and the open source movement, Perens turned to current events. He talked a little bit about how many companies doing Linux-related business suffer from multiple personality disorder. On the one hand companies like HP are looking to push Linux and are trying to embrace Linux and do the right thing for the Linux community. On the other hand, these companies have to maintain relationships with companies like SCO and Microsoft and participate in groups like CompTIA that actively work against open source. Perens cautions the community to pay attention to everything a company does, not just its support for open source.
We know that both Hewlett-Packard and the other members of CompTIA were
sponsors of the so-called Software Choice Initiative, which works
against open source. So, it's important to watch our friends.
Perens also noted that the next likely legal attack against open source would be via software patents, and said he thinks its unlikely that corporations like HP or IBM would help the community in that event. Though Perens says he hasn't made up his mind yet, he indicated he was thinking seriously about trying to form a community-driven answer to Red Hat's enterprise products.
I'm wondering if it's time for a grass-roots enterprise Linux, and
the way I figured I would do this... is first of all take Debian,
why is there a Fedora
project when there's Debian, a ten-year-old project with all its
policies done...with over a thousand developers? That is what the
Fedora project should be. Take that, and get together the
community of enterprise users who depend on Linux and really want a
zero-cost enterprise distribution.
After the talk, we caught up with Bruce for a few minutes one-on-one to ask about issues not covered during his talk, and to get further information on the grass-roots enterprise Linux effort. The first question was about the disagreement between the Free Software Foundation and the Debian Project over the GNU Free Documentation License (FDL). Perens has helped mediate between the groups, and says that they're on their way to working it out.
I wanted to take the emotion out of the whole thing, and it looks like
we're succeeding at that. I'm not tremendously happy to have coverage of
Debian and FSF bickering, we have a lot more important things going on.
I think that it's going to take care of itself more or less now. You probably will have some conference calls that are exciting... I'm not asking either organization to compromise with each other, I'm asking each organization to follow their own rules. I feel that it's not permissible for Debian to compromise its ethos for FSF nor is it permissible for the FSF to compromise its ethos for Debian and resolution of this issue does not call for either. Perens also clarified his thoughts on a possible "grass-roots" enterprise-ready Linux distribution:
It's something I'm still thinking about. I think I will go ahead and do
a call for people to work on it. Obviously I'm open for people who want
to discuss it. The project is not yet announced. I really debated this
in my own head for weeks now, and part of the reason was that, I feel
that it's a big personal expense to me to do any large project. On the
other side every open source thing I've ever done has paid back much
more than I've put into it... I feel that I must participate because I'm
one of the few people in the community with the cachet to pull this off,
who can talk to all the people on the executive side and all the people
on the community side and has reasonable credibility with both of them.
That doesn't mean I have to run it, definitely doesn't mean I want to be
its CTO, it does mean I would be evangelizing it publically for quite
some time.
I'm thinking about whether it is time for the community... to provide directly a Linux distro certified to LSB and to proprietary software providers that are willing to do so, guaranteed to be free software and free beer, free speech and free beer. A certified distribution that is zero cost, free software... and I'm convinced that creating a Linux distribution is a expense-sharing system rather than a profit-making system, even Red Hat now admits this as they attempt to offload production of their distribution to the community. We also asked Perens how he felt about companies that use open source software, but do not contribute substantially to the projects they use.
I have a scale for commercial collaborators with the community. It has
four points. It runs benefactor, partner, user, parasite. Benefactor:
NASA's a great example. They funded most of Linux's Ethernet drivers at
one time. At that time they were not able yet to make extensive use of
Linux, now they are. They put in more than they got out. Most companies
would not want to be benefactors, it looks bad to your stockholders.
Partner is what companies should be if they expect the cooperation of the free software community. At Hewlett-Packard, we could not get them to help us with the IA-64 kernel until we made the printers work. Very good lesson for companies, we put out 60 printer drivers on Linux because of that. User is a company that makes use of Linux and open source that complies with the licensing, but does not make any contribution unless they just can't avoid it. The usual GPL. I put Linksys in the user category if they finish resolving the issues they're working on with the FSF right now. Linksys is a division of Cisco, a very big company, that's important. Parasite, SCO comes to mind. They're making fraudulent claims to get value out of the Linux and open source community by kiting their stock and you can quote me on "fraudulent," "libelous," "slanderous," no problem with that. Other parasites, well who sold Linksys and Cisco that wireless access point? A chip company with a "B"... a number of engineering companies that seem to be in Taiwan and Korea, transfered intellectual property that was not theirs to Linksys and Cisco, in ways that did not comply with the licensing, leaving these companies whose goodwill we want out of compliance with our licenses and they don't know how to resolve the problem. So I don't like it because those Taiwanese or Korean companies made us enemies with Cisco when we want those guys to put Linux in their next product, we just want them to comply with the licenses and they should have been given full directions for doing so when they bought those WAP designs. Finally, we asked Perens if he had any thoughts on Eric Raymond's prediction that Sun is doomed.
Yeah, I wish Eric hadn't written that, actually. At least not quite the
way he wrote it, because first of all not having worked at HP as I have,
Eric doesn't understand how long a company can run on a legacy product
which is an extremely long time. And, secondarily, I think Eric was
angered by things Sun has been saying about Linux not belonging in the
data center and Sun's explicit collaboration with SCO spreading FUD.
However, Sun also helps us. Remember what I said about corporate
multiple personality disorders. They've done $70 million dollar investment in
OpenOffice, and I don't see where it paid off for them. They bungled the
strategic aspect of it, they need help with it, but it was a very large
contribution to Linux and open source. So, first of all, Sun's not going
away, they're not dying. If anything, they'll be acquired. They're still
a company with some value, and obviously their price is becoming more
attractive. Who will acquire them? I think it's either Microsoft or IBM.
We thank Bruce for taking the time to talk with us. (Log in to post comments)
free "enterprise" Linux distro Posted Oct 9, 2003 6:27 UTC (Thu) by snitm (guest, #4031) [Link] I work for a company that will likely be interested in getting involved in the development of this free enterprise distro. This is right inline with what we've been thinking is the ideal solution for our Linux distro needs.How many of you think this is something that would gain traction in the community? The 2 problems I see with it being based on debian is: Aside from that debian is ideal for such a distro.
free "enterprise" Linux distro Posted Oct 9, 2003 7:12 UTC (Thu) by stuart (subscriber, #623) [Link] um Debian does have native rpm support. apt-get install alienStu.
free "enterprise" Linux distro Posted Oct 9, 2003 10:27 UTC (Thu) by zonker (subscriber, #7867) [Link] I wouldn't call Alien "native" support for RPMs -- IIRC, there are a few steps required to convert an RPM to a Debian package for installation. Alien if fine for a single user converting a few RPMs for use on their Debian box, but it's not what I'd want to use for managing a large number of packages in an enterprise environment. There's also the issue of differences between systems -- so dependencies and such may be somewhat different, and the package converted from RPM to Debian format may not be used interchangeably.
free "enterprise" Linux distro Posted Oct 10, 2003 2:52 UTC (Fri) by Luyseyal (guest, #15693) [Link] Apt supports RPMs and dependency resolution natively. As long as they're not some shitty third party RPMs, they will work fine with Apt.-l
free "enterprise" Linux distro Posted Oct 11, 2003 17:42 UTC (Sat) by piman (subscriber, #8957) [Link] The 'differences between systems' is what the LSB fixes, so LSB-compliant RPMs should all convert file with alien. Otherwise it's either a bug in alien, or in the RPM.
free "enterprise" Linux distro Posted Oct 9, 2003 10:45 UTC (Thu) by tomsi (subscriber, #2306) [Link] One import thing that Debian is missing is a decent installation program - like SuSE/RedHat/Mandrake et.al.This is needed, because people who aren't used to linux will probably balk at the current install routines. Other important issues. Just my 2 cents. Tom
free "enterprise" Linux distro Posted Oct 9, 2003 13:33 UTC (Thu) by smoogen (subscriber, #97) [Link] I have to agree with that.. I did an install recently of the stable product and while it was better than it was 2-3 years ago.. it was not what I could sell to management. I am not talking about pretty pictures etc etc. I am talking about putting a floppy/cdrom into 20 machines and having them installed and updated in 15 minutes.Now while this is probably possible with Debian.. it wasnt as easy to figure out as the Red Hat install.
free "enterprise" Linux distro Posted Oct 9, 2003 14:20 UTC (Thu) by snitm (guest, #4031) [Link] with regard to debian stable needing to be more up-to-date (I agree); but I'd imagine that debian's _unstable_ is likely more stable than alot of what we'll see from RHEL3.0 in the initial release. RedHat hasn't won me over with their distribution quality/stability. That said, I hope I'm pleasantly surprised with RHEL3.0 because ultimately I'm forced to support it ;)
free "enterprise" Linux distro Posted Oct 9, 2003 22:01 UTC (Thu) by JoeBuck (subscriber, #2330) [Link] Even Debian unstable is rather backward compared to the leading GNU/Linux distributions. It has Gnome 2.2 (not 2.4), Xfree86 4.2.1 (not 4.3), etc. At least KDE is only one point release behind, and KDE 3.1.4 only shipped in September so this is good. But given the lengthy process to get a release out, I fear that sarge will be a year behind the competition on the day it ships. Debian seems to do very well with packages that don't have complex dependencies, packaging nearly everything under the sun. But they seem to be having far more trouble with X and the big desktops. Now, Debian labors under some handicaps: the fact that they support 11 architectures often means that they have to make packages work on oddball hardware that the upstream developers haven't seen.
Decent installer non-issue for enterprise distribution? Posted Oct 9, 2003 14:40 UTC (Thu) by Max.Hyre (subscriber, #1054) [Link] I can see how you'd want a gee-whiz installer if you were a home user, or a business owner wanting to do the work yourself. But if this is really meant to be an ``Enterprise'' (as in at least hundreds of desktops plus the servers needed to support them and the corporate workload), I'd expect the support staff to be perfectly comfortable with debconf. Set up a server, a firewall, and a desktop, and clone as needed. Once that's complete, apt-get should lower the maintenance level to well below that needed by RPMs. And if I were running a corporate server, I'd be perfectly happy with `stable', 'cause it sure is, and that's what I'd want. If an `unstable' package is the only one meeting a requirement, then install it, but if the servers run smoothly and do what's asked of them, what do I care what level they're running at? Certification (especially Oracle and their ilk) is another question, but if they certified against `stable' we have yet another reason to stay put until necessity calls.
Decent installer non-issue for enterprise distribution? Posted Oct 9, 2003 17:57 UTC (Thu) by tomsi (subscriber, #2306) [Link] I can understand that large organizations can cope with debconf; but there are other markeds for an Enterprise level servers, eg. the SOHO market. Those markets needs an easy to use installer with sensible package choices.I have never used apt-get myself, but I must agree that it sounds like a good choice for the day-to-day maintenance needs. I think that one of the reasons that Oracle et.al. don't certifiy against "stable" is that it is too conservative (IMHO).
Decent installer non-issue for enterprise distribution? Posted Oct 11, 2003 17:45 UTC (Sat) by piman (subscriber, #8957) [Link] Oracle doesn't certify against stable because to have Oracle certify you, they need support staff from your company in their call centers. You, the OS vendor, have to pay for this. Even if Debian had the money, it would be a violation of the Social Contract to spend it on overpriced support staff for a proprietary company.
free "enterprise" Linux distro Posted Oct 9, 2003 14:17 UTC (Thu) by wolfrider (guest, #3105) [Link] RPM is overrated, especially when rpm-based distros are including .deb support package-getters like "yum".
free "enterprise" Linux distro Posted Oct 9, 2003 14:30 UTC (Thu) by seyman (subscriber, #1172) [Link] What does yum have to do with .deb packages?
free "enterprise" Linux distro Posted Oct 9, 2003 14:32 UTC (Thu) by snitm (guest, #4031) [Link] yep, RPM pretty much sucks.... unfortunately its the chosen package format/manager of the LSB.Also, yum and apt-get (for rpm) can't hold a candle to Debian's apt-get w/ dpkg/deb on the backend. Debian packaging allows for much more package sophistication, BUT attaining that level of debian packaging sophistication has a sizeable learning curve that will likely impose some serious hurdles and doubt in the minds of existing rpm-based distribution software vendors and enterprise customers. Its unfortunate really, but Debian has generally prided itself on making aspiring debian developers run the deb packaging guantlet in order to prove they've got the required deb-fu. That's something that'll have to be lessened; possibly by leveraging some of the build systems that are coming into light from developers in the debian community.
free "enterprise" Linux distro Posted Oct 9, 2003 14:52 UTC (Thu) by Klavs (subscriber, #10563) [Link] I was thinking that Distro such as Gentoo would be good for such a project.Gentoo has ofcourse it's Portage - which supports packages and all that. It is definetely a long way behind Debian's dpkg system in that it doesn't do dependencies very quickly or very well yet. My reason for pointing to this distro, is that it fully supports dpkg and rpm packages on the same system. Only thing needed for this to work flawlessly, is to "tell" the different packages systems which dependencies are actually covered by another packaging system. If this was worked out - you would have a VERY flexible system, with support for whatever packages you wanted. The cool thing about using portage (the Gentoo package system) is that you can have your own versions of certain packages (not all of them - there is GRP - and there will be a continously updated, fully-binary, set of packages soon too), which means that people needing this - and many people do, can have this satisfied without having to write their own packages (requires a great deal of knowledge IMHO - I do it for customers - but it also means they have to maintain this, which also requires knowledge), and this also means that the system still can watch out what packages needs updating, and the only extra work for them (as opose to using the binary package) is a little extra compiling work - but NOTHING more - they don't need to understand anything - and the usual system security stuff can notice the package needs updating, and the package system can fully un-install their "special/custom" package. I'll stop myself, before I keep blurting out my ideas, but I hope you can see what I mean :)
free "enterprise" Linux distro Posted Oct 9, 2003 18:07 UTC (Thu) by snitm (guest, #4031) [Link] The level of control over gentoo that Daniel Robbins has jockied to maintain is SCARY, e.g.: the portage2 spat or the embedded gentoo brouhaha.Unilateral decisions, or abuse of power, in a project as big as Gentoo is hugely detrimental to the moral of all involved in its development. I doubt many companies would be willing to give it a go with Gentoo based on Daniel Robbins' track record. To be fair, maybe I'm missing Daniel's point of view... *shrug*
An Evening with Bruce Perens Posted Oct 9, 2003 11:06 UTC (Thu) by deatrich (subscriber, #25) [Link] Bruce said:> I'm wondering if it's time for a grass-roots enterprise Linux, and the way > I figured I would do this... is first of all take Debian, why is there a > Fedora project when there's Debian, a ten-year-old project with all its > policies done...with over a thousand developers?
Because, in part, it is the open-source way :o) People work on what please
An Evening with Bruce Perens Posted Oct 9, 2003 17:58 UTC (Thu) by zone (guest, #3633) [Link] >> why is there a Fedora project when there's DebianThis is obvious the question that occured to me when I read the Fedora annoucement. As far as I can tell, Fedora and Debian fill identical roles. Take Fedora's objectives statement, remove the RedHat-isms and a few minor points, and you've got Debian's Social Contract. If Fedora is going to fill any gaps Debian may not have filled yet, they'll need to re-package the vast majority of 10,000 packages, establish a Constitution in order to mediate disputes, and create a Policy Manual to set the technical tone and policies for the project. I have no doubt it can be done, and that Fedora can co-exist with Debian. The question is, why bother? > Because, in part, it is the open-source way :o) Competition is part of the Open Source way, sure, but reducing duplicated effort is a very big part as well. As well, competition at the application level is vastly more important than competition at the distribution level. After all, if you spend much time administering the distribution itself, it's probably not fulfilling its mission: to facilitate the use of the kernel and applications, which are what you're really interested in.
An Evening with Bruce Perens Posted Oct 9, 2003 19:13 UTC (Thu) by Per_Bothner (subscriber, #7375) [Link] > As far as I can tell, Fedora and Debian fill identical roles.However, Fedora is a compatible extension of Red Hat 9. Debian is not. This point is critical to both Red Hat (who wants to use Fedora as the base for their Enterprise products), their customers (who want stability and compatibility), and anyone else who wants Red Hat Linux. Using Debian is not an option. On the other hand, perhaps Fedora and Debian could/should merge at some point in the future, if they have develop similar comunities and philosophies, but that requires compromise on a number of technical issues. I think that is possible, and perhaps desirable, but there is a lot work needed first.
An Evening with Bruce Perens Posted Oct 9, 2003 22:08 UTC (Thu) by JoeBuck (subscriber, #2330) [Link] Fedora also has newer packages than Debian unstable does, though not as many. Gnome 2.4, Xfree86 4.3, kernels with a number of interesting 2.6 features backported, etc.
An Evening with Bruce Perens Posted Oct 9, 2003 17:31 UTC (Thu) by arcticwolf (guest, #8341) [Link] I'm shuddering at the thought of Microsoft acquiring Sun.
Debian and Enterprise Linux Posted Oct 9, 2003 18:31 UTC (Thu) by X-Nc (guest, #1661) [Link] While I completely agree that Debian is one of the best distros in existance, there are some fundimental "problems" with using it as the foundation of an Enterprise Solution.First, it's development cycle is, paradoxicly enough, both to slow and to fast. The "stable" tree is to slow WRT adopting things that are newer which businesses will likely need. The unstable tree(s) are, well, to unstable for a business to use. I know that they really are what we would call stable, but companies have a different POV on these things. Second, the deb package format is just not going to fly. Before everyone starts a debate (flame war) on this, the deb format is very good and technicaly excellent. The problem with it is that it stands alone. If a company has a choice of using what is percieved as a "standard" or something that is percieved as "radical" it's a no brainer which they'll pick. Remember, apt is not part of the actual package format but a utility for manipulating it. And it works just as well with rpm as with deb. Third, the installation and configuration of Debian are not easy enough to use for the scope of businesses. Even Solaris and AIX are much better and easier to use from this perspective. It seems to me that if you wanted to make a Free Enterprise distribution you could start with the srpm's from RH AS or ES. Take them, filter out anything that isn't as "free" as defined by whatever this proposed Free Enterprise distro decided to do, and build from there. At least that's an idea.
Debian and Enterprise Linux Posted Oct 10, 2003 3:20 UTC (Fri) by bajw (guest, #11712) [Link] While I am a stock holder in Red Hat, I mostly use Debian, though there are some problems there as well. For example, Debian is not an easy setup on a Mac even with a 2.4 kernel, due to the setup of a USB mouse, among other things. This type of issue will keep businesses away. There are other examples, such as newer hardware, even on Intel PCs. Look at the binary-only drivers for nVidia cards or other stuff in the Intel machines. Of course, Red Hat doesn't support Mac hardware at all and has the same Intel limits, so is certainly not better.But it is still true that RPM doesn't come close to apt-get for package management, even though either system is still rather poor for long term, continuing updating. In either system, the pain caused by an upgrade is usually solved/prevented by keeping all data in the (separate) /home partition, and doing a complete forklift upgrade by reinstalling to a newer system that just "comes with" all the desired hardware support and reinstalling the /home from backup instead of trying to keep up with knowing enough about the system to upgrade the "right way" in pieces. Even so, this is way too hard to do for most business users, due to the need to recreate each config file or other customization. Yes, the Debian way is better, but less "standard" (whatever that means), but it still doesn't make for easy computer replacement for an individual user in a business setting, or even simple hardware replacement/upgrade on a single machine, especially when the support for that hardware is only available in the newer kernels or other software not in "stable". There seems to be room for yet another distro, or a merger between Debian and Fedora, or some other solution to these limits that are too hard to deal with for most businesses, and will prevent their adoption of Linux and all the perceived troubles that would cause in the lives of the IT staff who would implement the change, and the managers who fear Free Software because of not seeing how to make it fit into their business views, which include a (rather fanatical) support of "Free Interprise" and all the patent/trademark/copyright things that implies to them. Some of it is a peception problem, but some of it is real. The combination allows the closed source, "business-oriented" suppliers to maintain a perceived advantage over the Free Software stuff.
Bruce: What's wrong with RedHat? Posted Oct 10, 2003 0:02 UTC (Fri) by pjhacnau (subscriber, #4223) [Link] I've seen Bruce Perens say elsewhere that RedHat is doing something wrong (spirit of OSS) with its Enterprise products. Presumably this is a follow-on from that. But I've never seen stated what the problem is. I could probably search around and find it stated somewhere, but I'm slack and I've seen Bruce post comments here occasionally. So Bruce: If you see this could you elaborate your problem with what RedHat is doing? Or if anyone else can point me to where he's already done that?
Bruce: What's wrong with RedHat? Posted Oct 11, 2003 16:02 UTC (Sat) by BrucePerens (guest, #2510) [Link] The most important things that a user-driven distribution can provide over RHAS are that the free version will be the certified one, that there won't be a lock on support information, and that it won't be dominated by one company. I am having talks with sponsors now. You'll hear from me in a few weeks.Bruce
Enterprise & grass roots won't work Posted Oct 13, 2003 16:01 UTC (Mon) by mwilck (guest, #1966) [Link] I doubt that enterprises will accept a "grass roots" enterprise Linux, because:1) managers only trust other enterprises, not grass-roots organizations. 2) managers want warranties and enterprise level ("24/7, errors fixed in < 2h") support. 3) For many software and hardware vendors, supporting (i.e. thoroughly testing and certifying) two distributions with all their security updates is more than they'd like to do. See how little commercial support Mandrake Corporate Server is getting compared to RedHat and UL. The vendors concentrate on the big ones, end of story. I am not saying these are good reasons, but it's the way the bosses think. That said, I do wish Bruce success in this project (perhaps my pessimism proves wrong). Technically, the goal should not be too dificult to achieve. It is amazing how little technical value the commercial enterprise Linux offerings have over the "normal" Linux distributions.
|
|||||||||||||