I was at a business looking to switch from HPUX/ Solaris to RHEL (Red Hat Enterprise Linux). We had some HPUX realtime(ish) extensions and used IPC heavily and really used the scheduler/Processor set functionality in HPUX. The scheduling was important, so transitioning wasn't going to be easy (endian issues aside).
They sent a bunch of us to a week long Linux Internals course at Red Hat. Really excellent class, knowledgeable instructor (turned us onto fedora, linux weekly news (https://lwn.net) and centos before red hat partnered with them). When the class wasn't exactly what we expected the instructor took the last day to go over some of the processor scheduling/ real time extension stuff we needed to know (My company's employees were the only one taking the class). The OS transition was shelved for a bit and I ended up leaving that company, but the course really changed my mind about that company.
Are there any open source companies that make even 1B a year? The reason I ask is because anyone with a model of "open source our core product" makes me concerned for their survival. It seems like a tough business model and Redhat is the only (moderate) success I can think of.
When your biggest competitor is yourself at a $0 price point, how do you compete?
I'll try to keep it short this time :-). Michael Tiemmann was one of the founders of Cygnus software, which was started on $6K and was sold to Red Hat for $600M. Tiemmann went on to become CTO of Red Hat and is currently (I think) VP of Open Source Affairs.
The key quotes from this link. When looking at the GNU Manifesto, "Suffice it to say that on the surface, it read like a socialist polemic, but I saw something different. I saw a business plan in disguise."
Later he goes on to say: "I would explain how freedom to share would lead to greater innovation at lower cost, greater economies of scale through more open standards, etc., and people would universally respond 'It's a great idea, but it will never work, because nobody is going to pay money for free software.' After two years of polishing my rhetoric, refining my arguments, and delivering my messages to people who paid for me to fly all over the world, I never got farther than 'It's a great idea, but . . .,' when I had my second insight: if everybody thinks it's a great idea, it probably is, and if nobody thinks it will work, I'll have no competition!"
IMHO opinion, Red Hat's best idea was acquiring Cygnus and buying into Tiemmann's ideas. The result is what you see now.
The thing that fascinates me is that Tiemmann still has no competition because nobody thinks it will work.
Having worked for a company that followed his model (Sendmail) I still say it doesn't work. I think Cygnus was a fluke. Also, assuming it was sold for 10x revenue, which is pretty standard, that means the revenue was only $60M a year. That's tiny for an enterprise software company.
Actually, IIRC revenue was $130M per quarter. In fact, after Cygnus was acquired the division it became drove revenue in Red Hat for a long time. It's been a very long time since I looked at it, but the records are public so if you are interested you could find out.
I'm curious about Sendmail. What was the business model? Cygnus's business model was essentially custom development. They did have a proprietary product for a short time, but as one of the terms of sale of the company they ditched it.
Just looking at sendmail now, it appears they were trying to make money dual licensing. I agree that this will never work.
Sendmail's original model was to get the best Sendmail consultants in the world under one roof and then deploy them building custom solutions.
Sadly a new CEO came in and declared that "the margins are in boxed software, so that's what we're going to focus on" and then fired all the consultants.
So maybe you're right, maybe it would have worked if they had stuck to the original plan.
But even then, it still wasn't a path to a huge business. 500M a year in revenue is definitely amazing and I would kill to own such a business, but it's not "rocket ship" successful, which is what most of the folks with this model are aiming for.
That's very unfortunate. I'd love to see someone else try that model, but I suspect that you are right that many people have a gold rush mentality. Getting paid for doing work is not as good as getting paid many times for having done some work in past. I've had a similar experience with a couple of startups I've worked on. You can show them how the custom development has high demand and is generating revenue, but they want to sell the box and gamble on getting their "multiples". From a certain perspective I can understand it. According to my link, Cygnus was limited in growth because they just couldn't hire people fast enough.
You sell services rather than products. Businesses outsource all the time, and why not outsource custom work to the same people who make the product you're using?
You're not going to be in the $1B space selling services, but neither are you at risk of extinction.
The problem with this is: billable hours are what make you money, not working in the product.
Juggling those becomes very difficult.
I've been following this space for years, and there are no easy answers.
I think the model where the 'secret sauce' is proprietary, and you give away the infrastructure, like, say, Ruby on Rails, seems to work ok. Produces some good open source software and is still an 'easy'/straightforward business model, because you're selling a product.
The problem with this is: billable hours are what make you money, not working in the product.
The two are not incompatible. Most of our new features are paid for by our clients, some of which we then include into our main product. Maintenance comes from other sources of revenue, though - hosting, for example, since many companies would rather not deal with that, even if they could.
The last time I worked at a software company they were trying to plump up the license $/services $ metric, because while it's easier to book services contracts they also increase your costs. Software companies only make runaway profits when they sell the same thing to a bunch of different customers.
That said, it's not like you can't make money selling services.
The OSS trade-off is market penetration vs revenue from a business perspective. It's successful if the free marketing and adoption is worth the loss of revenue.
Both RedHat and SUSE's competition in the 90's were proprietary Unix (e.g. SunOS etc) options. Linux commodified "low-end Unix-like OS" in conjunction with Intel processors. As a thought experiment, imagine them trying to compete against the big UNIX hardware vendors as a pure OS vendor: it would have been impossible, and there are some parallels in the outcomes for BeOS and NeXT. Both RH/SUSE were able to widely penetrate through "cheap distribution", getting their product into the hands of users who then preferred to use Linux - plus it was cheaper.
Strictly speaking Open Source is not failed business model because it simply isn't a business model at all. It's a model for licensing software, not how to make money. But, I know what you mean, it does add another level of complexity.
However, there are ways to run a model that has many of the advantages while also keeping value in the commercial aspects. For RH/SUSE they have two versions - the free stuff, and their certified distribution. The basic version of that is that their paying customers are primarily slow-moving corporates and the military. While people say they sell "support", a key part of the package is hardware and software certification - they're selling compatibility and stability and certainty. If you're in IT in a Fortune 100 and you want to run SAP, even if your tech team could run Fedora, SAP isn't certified on it, and if things do go wrong do you want to be in that conversation?
Good. Some use Linux, and barely contribute back. Getting RHEL subscription is a good way to do it, because RedHat does a lot of work on improving Linux.
They pay me to continue working on my desktop Linux project (from the bay area no less) and have in no way impeded my personal direction for the project. Quite the opposite, I now have a team of people willing to help me move faster by solving external problems as I reach them.
So the only negative thing I can say about redhat was that whilst dealing with implementing Openstack for them, they surely had become a behemoth of a company... meaning that they were at one point the scrappy up-start we all wanted to beat IBM/HP etc... but they had become just as "enterprisey" as the others.
Obviously there is nothing wrong with that as their revenue shows, its just that the internal corporate DNA changes to a certain degree.
A massive chunk of our engineers are (globally) remote. A lot of communication happens via mailing lists, many of which are the public upstream project lists. There is also a lot of use of internal IRC servers for more real time communication.
Outside of that there is also a fair bit of freedom/latitude for individual teams to do what works for them.
I hear there's movement away from IRC and onto Slack, which is apparently contentious internally. Interesting to hear nonetheless given it isn't FLOSS.
Good for RH! AFAIR they abandoned the freely-downloadable Red Hat Linux, and started charging for the pre-compiled versions of Red Hat Enterprise Linux (RHEL), and this is when they started making money to be where they are today. Didn't they have a falling out with Linus?
That was really just rebranding. Red hat Linux split into Fedora (free) and RHEL (for pay). Even so, the source for all the rpms for both distressing are available, and CentOS has been using that to build their RHEL clone for years, and has always had an amicable relationship with RedHat as far as I know.
He had one of his tedious, high-profile, foul-mouthed rants at a RH kernel dev when Linus didn't understand the dev's patches. Is that what you're thinking of?
They did give Linus a boatload of shares around the time they were going public.
Businesses love it because every major release is like Ubuntu's LTS releases. RHEL 4 (released in 2004) is still in extended support!
I assume their support is good also, but I haven't used that.
For me though, I have a bunch of servers running CentOS just because it's extremely solid and stable. The price you pay is that all your packages are a little old, but there are repos to get new software, so I can be on the newest Postgres and Node etc.
Can you point me to the Debian developer who will supply me a kernel patch, on my time line, for a feature I need?
Can you point me to the Debian developer who will give me follow the sun support on a problem that could be with the J2EE container, the OS, the virt layer, or the hardware, and has support arrangements with the vendors of all those things so they don't just sit around finger pointing?
Can you point me at the J2EE container Canonical supply?
a. Yes, Canonical will provide you with a patch to an issue in Ubuntu the same as RedHat.
b. Yes Canonical has follow the sun support for the "Main" section of Ubuntu. This includes the OS, the virt layer and the hardware along with relationships with the hardware vendors.
c. No Ubuntu doesn't do enterprise J2EE
The only open source solution is TomCat and as you know they won't co-operate with other distributions.
If you have hundreds of HP Proliant servers, all with the latest firmware patches and some level of support, and they all run a supported version of RHEL, with some level of support, and once a week a different machine crashes and dumps the kernel core - if you have RHEL (and HP) support, you have a way to escalate the issue. They have kernel programmers on staff who know what errors might be in the kernel code that talks to, say, host bus adapters. What does a normal corporation do in this circumstance if they are running Debian servers?
Insofar as competition in this space, SLES has been popular in heterogenous environments (a lot of Windows servers). Canonical and Ubuntu are newer competitors on the scene. Companies have been using Red Hat for 20 years, they are pretty good at their jobs.
Also, RHEL is a flagship but Red Hat sells support for other products like JBoss.
RHEL is 800 per server. If you have hundreds of these servers, you could pay a developer 80k to sit around learning kernel development on the offhand chance this comes up.
You could, but that's going to be the unloaded cost of a developer, add another 80k for benefits, tax, and everything else that comes with an employee. Even after that you've got a single kernel developer who might have learnt enough about the particular part of the kernel you're having problems with. Assuming its the kernel.
Either that, or pay 80k to Red Hat, who have dozens of kernel developers one of whom probably knows about your issue, and if it turns out not to be the kernel probably have someone on staff that can investigate that as well. Worst case they know who to contact upstream to dig deeper into your issue.
And the good thing is that companies are free to do that.
Take Google, for example. They don't pay Red Hat because they have the code and they have amazingly smart kernel developers. But that's (surprisingly) still good for Red Hat, since Red Hat benefits from the contributions to the kernel that Google makes. It's good for all users of Linux.
What's so good about RHEL/fedora vs debian or ubuntu server, say?
It's mostly just a matter of personal taste / preference. I use both Ubuntu and Fedora/CentOS and there's not really any huge functional differences. For a long time, years ago, people argued that apt was a much better tool than yum, which may have been true at one time. But for quite some time now I've found yum to be equal to or superior to apt. Of course, Fedora has moved to dnf now, where dnf is a more modern, drop-in replacement (mostly anyway) for yum.
For me, I tend to stick mostly with the RH derived distros just out of familiarity. I've been using Red Hat Linux (and derivatives) since RH 6.2 was out, and I know it better than I know Debian or Ubuntu or Slackware or whatever.
I've always been a Debian/Ubuntu guy, but the Redhat folks are good people and have continued to dedicate a lot of resources to open source software, through thick and thin.
Sure business can pay for Debian enterprise support too, but its slightly different mindset when the whole ecosystem was made for business and enterprise.
When you think about it, it's crazy it took them this long to make $2B in revenue. It took them 23 years since it was founded in 1993. Facebook hit $2B in revenue 6 years. Which is crazy when you look at the comparison and RH is the most successful open-source co. Open-source is a failed model for commercial success.
Another way to look at it is that proprietary software vendors drain massive amounts of money out of the world, far in excess of the value they generate.
Red Hat open sources all customer facing software they create. When Red Hat buys a company, it is expected that they will open source its software products. I think "open source company" is only apt.
With the acquisition of Ansible, we haven't seen Red Hat releasing the products (Ansible Tower). Is that just a case of they haven't got around to it, or is this a new strategy?
They will open source it, but then can't just publish all the code without reviewing it. What if there is some code there that they can't legally but into the public space? It's legalwork, that shit takes time
This is happening. For comparison, it took Red Hat 4 years to open source RHEV, because of (a) legal shit (b) having to rewrite a bunch of .NET code. But it did happen.
I think what sets Red Hat apart from companies like Microsoft, Google, Apple, etc. is that those do have a lot of open source code, and it's an important piece of their strategy, but their most critical code is not open source. And not just the most critical code, but most of their code is not open source.
Open sourcing some strategic things makes sense for them. But for Red Hat, open source itself is the core strategy, and they open source everything.
I meant attitude towards free software in general (i.e. stance on software patents and such). That divides those who can be called open source company, and those who can't. It's not a sufficient condition, but I'd say it's a requirement.
'open-source' means that the public can examine the code. If Windows and Office have publicly available code (being MS's core offerings), then yes, you could claim MS was an open-source company.
A company doing something on a fringe doesn't mean the company is defined by that fringe. Microsoft releasing a bit of open-source work doesn't make them an 'open-source' company any more than giving out freebies at a conference makes them a welfare agency.
Publicly visible source code is not enough. You need be able to modify and (freely) distribute changes to that source code.
When I can write my own patched version of Windows (A Windows distribution if you will) and people can download my modified version then it will be open source.
> You just described free software, not open source.
Someone hasn't read "the open source definition". But in either case, free software is what matters here. What you describe as "open source" is actually proprietary software.
Also, what software did mpeg describe? I can't see any such reference. And if you mistakenly meant my reference to Windows/Office, well, that was meant as a hypothetical.
What term would you use for "publicly visible, but not hackable-by-license" software as opposed to "publicly not visible, and not hackable-by-license"? The former satisfies many concerns with vendor software, in that you can examine the code yourself for bugs, vulnerabilities, or backdoors.
That link doesn't seem to have anything to do with what you said. That isn't Stallman's annotations -- he doesn't use the term "open source" and probably wouldn't want his words published on opensource.org. You probably wanted to post this: http://www.gnu.org/philosophy/free-software-for-freedom.en.h.... The main reason that the FSF and Richard Stallman distinguish free software and "open source" is because "open source" doesn't reference the moral underpinnings of the free software movement. It's just a set of 10 guidelines that appear to be entirely arbitrary (because they're secretly derived from the free software definition, but with more restrictions in some cases).
> What term would you use for "publicly visible, but not hackable-by-license" software as opposed to "publicly not visible, and not hackable-by-license"? The former satisfies many concerns with vendor software, in that you can examine the code yourself for bugs, vulnerabilities, or backdoors.
I would call both proprietary. The distinction is that they give you the code for the program as part of their proprietary software package. It doesn't make it non-proprietary if that's the distribution model they use.
No. Both sides of the FSF/OSI schism are the same page with respect to Modification and redistribution. Being able to see the source is not enough you have to be able to use the source in a meaningful way.
Same thing with an NDA. if you make source available under NDA that is not Open Source either.
The FSF acknowledge that "open source" is a bad term as it's easy to misunderstand for "code for which the source is available to the public". I completely agree that source under NDA is a different scenario.
And yes I know that OSI don't have an official trademark on the term Open Source and that the OSD isn't a de-jure standard. But I'll argue that it is the de-facto standard among knowledgeable people working in this space.
They sent a bunch of us to a week long Linux Internals course at Red Hat. Really excellent class, knowledgeable instructor (turned us onto fedora, linux weekly news (https://lwn.net) and centos before red hat partnered with them). When the class wasn't exactly what we expected the instructor took the last day to go over some of the processor scheduling/ real time extension stuff we needed to know (My company's employees were the only one taking the class). The OS transition was shelved for a bit and I ended up leaving that company, but the course really changed my mind about that company.
Good for them.