Why There Will Never Be Another RedHat: The Economics Of Open Source | Hacker News Reader

Why There Will Never Be Another RedHat: The Economics Of Open Source

techcrunch.com

59 pointskanche66 comments

It's an interesting worldview where growing a company to an $11bn market cap counts as only a 'lukewarm success'.

It's not super clear why Red Hat's market cap of "only" $11bn means that their business model doesn't work. It would seem to me that it works pretty well -- to the tune of $11bn.

I think it's more that the economics of RHT are skewed compared with the marketshare that they've got.

That is, Red Hat has basically won the enterprise datacenter Linux war. It's Red Hat Linux and Windows, at the cost of Solaris, AIX, HP-UX, SLES, and others.

And despite that crushing success, and having no worlds left to conquer in the OS space, their market cap is only $11b. Salesforce.com has a 12% market share, and they have a $38b market cap.

EDIT: just cleaned up "Linux" vs. "Red Hat Linux"

This is not a stable success. The advantage of RHEL over other posix-y flavors is 1) cost, and 2) standardization of server build and software packaging.

Solaris / AIX / HP-UX still get plenty of mileage in the legacy application (Oracle!) space. It's going to take the enterprise longer to migrate to systems like CoreOS, but RHEL doesn't have the kind of automatic lock-in that some of these legacy systems have.

As far as I can tell, RH is focusing revenue growth strategies on things like OpenStack, and similar product suites, so the company will survive. But by the time RHEL 7 comes out, what's it going to offer anyone in the datacenter? The days of the monolithic distro are numbered.

I'm not arguing it's a forever stable success, because the industry moves too fast for that. And yes, this is why RHT wants OpenStack and better diversification.

But speaking as someone who literally based their whole business on managing Oracle on *nix, Solaris/AIX/HP-UX are dead platforms. Over 80% of the Oracle servers our users are managing with our software are Linux. Of those, probably 80% of them are RHT, and 15% are Oracle Linux. SLES and CentOS barely register.

There is one more massive advantage that RHT offers vs. anyone like CoreOS - platform stability. Things like driver interfaces, library versioning, kernel versions all make it much easier for ISVs and corporate engineering teams to certify and manage RHEL releases. They're not a fast-moving target.

Over time that will become less relevant, I agree, but it's been a huge factor in their success.

I only meant to pick on the "RHEL and Windows" remark. The OS is only a small part of the datacenter.

I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category. Oracle the DB is one thing, but the sprawling ecosystem of non-database applications that Oracle has accumulated over the years aren't going to keep needing heavier distros. A lot of that stuff is moving to VM image instances anyway.

What a platform like CoreOS (or other ways of combining LXC with your hardware infrastructure) offers is separation of two major categories of concern. Driver interfaces and kernel versions are hardware sensitive. Application dependencies on specific library versions far less so.

Your vantage point makes sense, though. Databases are really the only application left in the enterprise that haven't seen this decoupling from hardware. But that doesn't bode well for the total number of Red Hat Linux installations.

I realize the OS is only a small part of the datacenter, but it's most of RHT's business, so that's what I focused on.

> I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category.

I have a hard time seeing big iron becoming anything other than the open systems equivalent of legacy mainframe.

Bear in mind that kernel interfaces aren't purely about hardware - it's about repeatable, predictable performance and behavior profiles. Watch a bunch of bank server platform engineers lose their shit when memory performance under pressure changes after a kernel patch.

And remember that in the traditional enterprise, it's not just about stripping down the OS, it's about all of the other tools that get used for performance, security, backup, monitoring, remote execution, change management, privilege escalation. I actually am working on some research on things like LXC at my day job, and one of the challenges is that containers are great until you want to run a bunch of stuff inside the container.

And my vantage point was at one time databases, and that's still the market segment I know the best. But I work on everything server and cloud-related, so most customers I talk to are just traditional enterprises.

Most people I know running servers don't use RH... they use debian or ubuntu.

My company (15k+ servers) uses ubuntu.

Tends to be the circles one runs in. For comparison, I don't have a single customer that I know personally that runs Ubuntu as a platform strategy, though I'm sure there's pockets of Ubuntu in many of them.

Our customers are big traditional enteprises, by and large. Banks, telcos, manufacturing, oil and gas, etc. They love RHT, or if they don't love it, they at least appreciate the stability and ISV support.

Really depends on the industry and company. My experience is that less venturous firms tend to run RHEL or CentOS, more clueful / adventurous ones Debian / Ubuntu, though mileage varies.

I sort of remember this was a talking point for red hat advocates way back in the early days, that it's better to have 50% of a tiny market than 5% of a huge one.

I agree completely. If open source doesn't require as much up front investment, then you don't need as much sales or market cap to justify good returns for your investors.

That aside, I do like the point of the original article. Build a scalable service on top of something open source, rather than T&M services.

"Build a scalable service on top of something open source, rather than T&M services."

37Signals/Basecamp are an interesting example of this. Rather than building a business around supporting Ruby on Rails, they've built products on top of the framework they created. It's certainly an outlier and I can't think of many other examples like them who both built a popular tool, but then just used it as a tool instead of trying to monetize it directly.

DHH wrote: "External, expected rewards diminish the intrinsic motivation of the fundraising open-source contributor. It risks transporting a community of peers into a transactional terminal. And that buyer-seller frame detracts from the magic that is peer-collaborators.

It also holds the threat of corrupting the community at large. We plant the seeds of discontent by selective monetary rewards."

http://david.heinemeierhansson.com/2013/the-perils-of-mixing...

I've certainly seen this recently in some open-source communities. The peer-to-peer camaraderie transitions to a dictator-to-underlings relationship and people start defecting. It's sad to see, but happens when an open-source company isn't vigilant in nurturing the community that got them where they are.

Kudos to those companies that are able to pull off maintaining a strong community while still staying focused. It's a rare thing to be able to do well and behind the scenes there's a lot of work that goes into that kind of nurturing.

He also completely ignores the notion of time. Compare the size of the companies relative to their founding date and your going to see the following:

VMware 1998 Microsoft 1975 Oracle 1977 Amazon 1994 RedHat 1993

At least two companies had a major head start and should be expected to be larger.

One doesn't sell software, it uses software (Amazon).

The remaining two are VMWare and Redhat and the success of either can be attributed to the birth of their markets.

The virtualization hardware mostly began with a commercial product and the commercial products were able to stay ahead by being better capitalized than any efforts to commoditize the product.

With the other market, server operating systems, Linux was already a community effort with several distros and you couldn't have exactly have even tried any other strategy than open source without completely alienating your early adopters and dying on the vine before getting to a minimum level of adopter.

So many shaky premises here that I barely know where to start.

1) Canonical. You could have made the exact same argument as this article 10 years ago and said that there wasn't room for another RedHat. Canonical arrived and proved that yes, there is room for another profitable Linux vendor. I know I use Ubuntu server and am considering support contracts, so there is definitely a market.

2a) So Redhat doesn't have the resources for development that Microsoft has and therefore will produce an inferior product? Have you seen Windows 8? Vista? Money does not guarantee success.

2b) Even without the same resources, the main Linux kernel has an enormous number of top-talent developers working for free. This is a serious economic advantage over traditional closed source companies.

3) Why on earth is the article comparing RedHat to Amazon? One is a Linux distribution vendor, the other is a global internet shopping mall.

I'm sure others can provide more reasons. Or even counterpoints; feel free to rebut ;)

Canonical is not profitable.

The parts that do what Red Hat do are; the overall lack of profitability is because Canonical is losing money on other market segments. (In other words, they're generating money to be invested in developing new products, exactly what the article says open source companies can't do.)

My bad. Perhaps I should have said sustainable. Canonical is not going to disappear. They may have unprofitable ventures, but ultimately they can cut it back to the core business and carry on going. Got to spend money to make money!

> Why on earth is the article comparing RedHat to Amazon?

I don't understand that and not because:

> One is a Linux distribution vendor, the other is a global internet shopping mall.

I don't understand that because Red Hat already does what Amazon does:

http://www.redhat.com/solutions/open-hybrid-cloud/

I don't know if Peter Levine knows that.

Yes, but the article is comparing the total Amazon revenue in the $70bn+ range to the RedHat ~$1.3bn. A completely misleading comparison, considering that Amazon don't actually break out their AWS figures in their financial reports.

Those charts are highly misleading and borderline duplicitous.

He shows estimated AWS revenue on [1], where he appears to be using the $3.8 billion revenue estimate. I regard that estimate as defensible, and a reasonable indicator of the size of the market.

[1] http://tctechcrunch2011.files.wordpress.com/2014/02/peterlev...

[2] http://www.zdnet.com/amazons-aws-3-8-billion-revenue-in-2013...

Yes, because "providing service and support for software running on someone else's hardware" is the same as "maintaining large datacenters, including hardware, power and cooling costs," and therefore let's just compare revenues without looking at costs.

If I was an investor, and I saw Redhat growing revenue at the rate AWS seems to be then I'd be buying as much Redhat stock as I could - even if their profits were depressed because of the investment they were making.

I'll quote the investment guidance from the link I posted above[1]:

Using our estimate of $3.8bn for 2013 AWS revenues, and applying a ~5x multiple based on the comps noted above, we arrive at a valuation of ~$19bn for the business on an EV/Sales basis (equating to ~$41/share of AMZN stock). Importantly, we believe this to be a conservative valuation multiple, as AWS revenues are growing much faster than any of the comps incorporated above. At an 8x valuation multiple, we estimate the AWS business could be worth $30bn as a stand-alone company, or ~$66/share.

Note that they value AWS as a $30bn company right now, almost triple Redhat's $11bn market cap.

From [2], we see that Amazon has has invested $4.6 billion in its business over the past year, including $1.4 billion to purchase the building that houses its Seattle headquarters.. Taking out that $1.4 billion, we get $3.2 billion, which almost matches estimates for AWS revenue.

So we see AWS is growing like mad, pretty much owns the whole cloud hosting category[3] and MIGHT NOT EVEN BE LOSING AMAZON MONEY, even while they are investing so much that no one else can compete with them on scale of their offerings already.

Take my money now.

[1] http://www.zdnet.com/amazons-aws-3-8-billion-revenue-in-2013...

[2] http://www.informationweek.com/cloud/infrastructure-as-a-ser...?

[3] http://readwrite.com/2013/08/21/gartner-aws-now-5-times-the-...

You have to remember how AWS started - Amazon selling their excess datacentre capacity in their offpeak downtime.

It really isn't possible to distinguish how much of AWS is a money maker for Amazon, because you can't separate the costs of running it from the costs of the Amazon business as a whole. For all we know, they could be selling the excess capacity at a massive loss, which is still justifiable for them because at least it recoups some of their everyday running costs.

Red Hat already does what Amazon does

No, they don't. You can't pay RedHat money and run software on their infrastructure.

Instead, Redhat is building a set of open source platforms, which they will sell support to you for.

Their IAAS platform is based on OpenStack, and their PAAS is OpenShift[1].

The problem isn't the software, it's that they are relying on support revenue for these too. As Levine points out, that isn't as profitable source of revenue as running your own cloud service, or selling a SaaS app.

I don't know if Peter Levine knows that

Levine was the the CEO of XenSource (who wrote the Xen hypervisor). Pretty sure he knows the cloud space fairly well.

[1] https://www.openshift.com/

"You can't pay RedHat money and run software on their infrastructure."

You definitely can.

https://www.openshift.com/products/pricing

>2b) Even without the same resources, the main Linux kernel has an enormous number of top-talent developers working for free. This is a serious economic advantage over traditional closed source companies.

I am not familiar with the actual org structure of Linux kernel contributors besides Linus being the leader but isn't it the case that most of the major contributors are in fact employed by companies (including Red Hat)?

In other words not devoting their time working for free...or am I off?

For RedHat it's free.

http://arstechnica.com/information-technology/2013/09/google...

If 10% are paid by RedHat, 90% aren't - therefore count as free to RedHat. Even then, the single largest group are non-aligned developers.

I agree Red Hat and pretty much every other contributor's result is a net gain. The point I was making is that the overall contribution isn't free. ie the folks working on it are paid by various companies. I believe all the major tech companies have a few devs that they basically loan out to the kernel development effort.

Yes, there's potential for a Tragedy of the Commons style problem to happen with kernel development, where each corporate developer suddenly decides to save money and not contribute, whilst letting everyone else pay. Eventually everyone would follow suit and nothing would get done.

Not exactly the same as the Tragedy of the Commons, but similar in the sense that it works best when everyone works together.

NB: It's also important to note that some kernel work is actually on drivers for specific products. eg. Intel's contributions are largely making sure that all their hardware will work in a Linux server/desktop, not necessarily the kind of general Linux ecosystem development that Canonical might do.

I don't see the tragedy of the commons concern here. Presumably, each company is contributing because they need/want something with the kernel that is not already being provided. Instead of keeping the development in house, which would involve maintaining a fork and almost certainty breaking copyright, they decide to contribute it back to the mainline kernel. At the point where no one wants to contribute, I would expect that the expected value for some companies of having a single employee work on the kernel is greater than the That this happens to benifit everyone else is just a side effect.

Many of the kernel developers are paid for their efforts by somebody, yes. Red Hat is the top single corporate employer, but their contributions are roughly 12%:

http://www.linuxfoundation.org/sites/main/files/publications...

From Red Hat's point of view, even though someone else is typically paying for the effort, code that's not from their developers is "free." (Well, Red Hat is also one of the top reviewers, so they're paying people to review the contributions of others, too. So it's a bit more complicated than that.)

Now, some of the companies on that list are in the same business as Red Hat (and some companies are in the same business but don't contribute significantly to the kernel, like Canonical). But many of them are in very different lines of business and view the Linux kernel as either a complement (that's why Intel is involved, for instance -- encouraging free operating systems to sell processors) or as an infrastructure piece for their products (yo, Google). So Red Hat is able to leverage the efforts of those companies to deliver a product for far, far fewer man hours than it would cost if they did all the work themselves.

Most contributions are from volunteers, but I do not know about contributors. I assume volunteering contributors are even more, because they are sort of the fly by crew.

Not sure of that...I may be wrong but I believe there is a karma type process in place for instance regarding who can contribute patches...I can't imagine there is much fly-by-night contribution...except perhaps bug reporting.

RedHat contributes ~10% of total development:

https://lwn.net/Articles/451243/

so from the point of view of RedHat a large part of kernel development is actually free. Not actually 90% though since they don't care about all changes.

Getting unrelated and often competing companies to cooperate in creating the infrastructure layer of software is a huge success case of open source.

Canonical arrived and proved that yes, there is room for another profitable Linux vendor.

Cananical only had $65.7 million in revenue in 2013, and made losses of $21.3 million[1].

I think that proves his point rather well.

So Redhat doesn't have the resources for development that Microsoft has and therefore will produce an inferior product? Have you seen Windows 8? Vista? Money does not guarantee success.

Microsoft's product development problems are well documented. Their money making capacity is equally as well known. Their last quarter broke revenue records for them[2], helped by growth in both enterprise and cloud.

Even without the same resources, the main Linux kernel has an enormous number of top-talent developers working for free. This is a serious economic advantage over traditional closed source companies.

There are a number of unpaid contributors to the kernel, but the vast majority are paid by companies. 83% of kernel 3.10 was funded by a range of companies.[3]

Why on earth is the article comparing RedHat to Amazon? One is a Linux distribution vendor, the other is a global internet shopping mall.

Are you trolling? The author posted a graph showing how AWS (Amazon web services) has more than double the revenue of RedHat.

That's why other operating system vendors are (eg, Microsoft) are pushing into the cloud space, and why RedHat's lack of a coherent story on cloud is a big problem for them.

[1] http://arstechnica.com/business/2014/01/ubuntu-maker-boosted...

[2] http://www.pcworld.com/article/2055455/record-first-quarter-...

[3] http://arstechnica.com/information-technology/2013/09/google...

A lot of Microsofts growth in recent years has been in enterprise tools, not operating systems.

Amazon do not break out their AWS revenue from their main revenue, hence one of the reasons it's a dodgy claim. Not to mention the fact that the article included the total Amazon revenue in the charts for no good reason. Likewise for Microsoft, who have multiple large revenue streams in the form of Office, etc...

Not to mention Oracle, who are a database company that dabble in Linux (I've never actually met anyone who uses "Unbreakable" linux - yes, yes, anecdotal I know).

When you remove the specious data points from the graph, you're left with RedHat on it's own. Even if you include the AWS estimate, you still have a much more reasonable graph that indicates RedHat competing in the same ballpark as AWS.

A lot of Microsofts growth in recent years has been in enterprise tools, not operating systems

That's exactly the point (?!)

I think you are looking at Redhat, and seeing them as a successful operating system vendor.

Levine is looking at them as an enterprise software company who haven't grown revenues and/or profits at anything like the rate that other enterprise software companies have.

He then sees that there are no other large software companies which have a business model similar to Redhat.

He uses those two pieces of evidence to argue that there are better business models for Open Source than what Redhat does.

You appear to be picking at things like if Redhar competes with AWS. That isn't his point at all - he's showing that AWS is an alternate business model which appears to have grown revenues much more successfully than Redhat.

Fair enough. But the entire point of the article is that "There will never be another RedHat".

It's the title, it's presumably the main point of the article - and it's ridiculously over-generalising.

RedHat as a company, works.

The market share is still largely with Microsoft, thus still available to be stolen away. Therefore, there is still room for a huge amount of competition in the OS vendor/support space. I can't see how anyone can argue against that.

I'll grant you that perhaps RedHat could be making more money by spreading out into different sectors like SaaS and PaaS, but that doesn't mean that there can't be another RedHat in the market.

I guess my main complaint is with how this point is argued. The charts are deeply misleading for the point that is trying to be made.

Please, don't start pushing SaaS into FOSS under the guise of profitability. Sure, I'll admit that the support model (RedHat) has it's issues and probably keeps profits lower than competing proprietary solutions, but lets not forget the third model: up front sales.

Here's a problem I run into frequently: I have a problem, but as a look for a software solution to that problem, I find 3 potential solutions. Of those solutions, one is a random sourceforge mish-mash that hasn't been maintained in years, and the other two are proprietary closed source solutions (which I try to avoid whenever possible. Why should I trust some random small company I've never dealt with to not be malicious or incompetent?)

What I really want is a good, up to date software that I can pay for, but is still open source. I think too many people have forgotten that open source doesn't mean that you have to release the code in public for all to see on github et al, it just simply means that you have to provide the source.

So unless you are a customer, you don't get access to the source (unless the customer then shares it... this is probably one of the few areas I think that needs more debate in the GPL vs MIT war.)

With all of the revelations of the NSA, I think any business that deals with sensitive information in any way has a responsibility to 1) move away from SaaS and cloud computing, and 2) move away from proprietary whenever it is possible to do so without negatively impacting business operations.

We need to reclaim control of our data, as users personally and in business, and the only way we reliably do that is with decentralized, locally controlled FOSS.

I have a small disagreement with the market cap/revenue argument: "its key point of failure is that the business model simply does not enable adequate funding of ongoing investments." Red Hat never intended to become as big as MS (they've repeatedly claimed their goal was to actually shrink the OS market by selling Linux cheaper) because they don't have to fund the entire development of Linux themselves.

they've repeatedly claimed their goal was to actually shrink the OS market by selling Linux cheaper

And they shrunk Sun right out of it (along with the other legacy Unix vendors for the most part).

I'd argue that it's Microsoft which is the anomaly.

Your post really ought to be more widely recognized.

The economics of open source software is fascinating.

I think the best model for open source is to be the "base of the pyramid", with companies adding a little bit of proprietary stuff that differentiates them at the top, and collaborating on the infrastructure below. Stuff like Ruby on Rails works this way, with Basecamp being the point of the pyramid, and Rails, Ruby, Linux, Mysql etc... being all the stuff it's built on.

If you start with GPL stuff (and in my experience most of the really good open source software is GPL), it's very difficult to sell the resulting proprietary product, as you must provide the source on request and you cannot prevent it from being shared; thus, anyone can build it themselves (some projects I've seen that use this model deliberately make their build process tricky, I suppose to encourage sales-by-laziness - Apple is the king of this with their essentially-unbuildable releases of Darwin).

No, I rather think the best model is the one where software is treated like a public resource; the knowledge commons. A lot more companies are interested in using software internally than in selling it, and many will already have in-house software developers to build their business-specific systems. It makes a lot of sense for companies to pool their resources on common software problems that they all need to be solved and aren't their core focus.

In a sense, your "base of the pyramid" is something of a special case for this model. Basecamp is software-as-a-service, and Ruby on Rails is utility software designed for the express purpose of creating software-as-a-services. It's to be expected that, as free software gains momentum, the first companies to benefit will be the ones that build applications that need to run on reliable, non-locked-in stacks, because a reliable, non-locked in stack is the first thing that free software needed to build. In the long term, though, I hope that really substantial "user-facing" software gains enough critical mass to become the de-facto standard - for example, Blender or GIMP. In such a scenario, any business that needed an off-the-shelf 3D editor or image editor but also developed in-house software (for example, Pixar) would find it in their interest to patch bugs and generally improve the software that they use, even though they don't sell it.

>In the long term, though, I hope that really substantial "user-facing" software gains enough critical mass to become the de-facto standard - for example, Blender or GIMP. In such a scenario, any business that needed an off-the-shelf 3D editor or image editor but also developed in-house software (for example, Pixar) would find it in their interest to patch bugs and generally improve the software that they use, even though they don't sell it.

I think this is the Achilles heel of free software. It works well for infrastructure software such as OS and middle-ware solutions but doesn't and cannot work well for end-user applications.

When you are working on the Linux kernel you only need to concern yourself with the correctness of your solution. When you build end-user applications it is an entirely different concern, you can have the best technical solution and still fail to address users needs. The moment you decide user needs are a priority, your cost is going to go up and more importantly the community strength is likely to falter because a good chunk are going to feel unrepresented.

Whether we like to admit it or not, software developers tend to build products for other developers, no matter how much they try the end result will end up appealing more to developers than the general population. This is illustrated by the weak show of Linux on the desktop, for years they've tried but not much has changed.

This also happened with smart phones, some might not remember but one of the first truly exciting mobile devices where the Nokia 770 & N800 internet tablets, they were awesome devices but alas were quickly supplanted once the iPhone made its debut. I think this shows that while those devices were awesome (full linux in your pocket, which geek wouldn't love that?)to us as developer they failed to address the general population's needs.

The most successful end-user open source application so far is Firefox and that is because Mozilla is basically a quasi business organization and are able to approach developing Firefox like a commercial product.

> Whether we like to admit it or not, software developers tend to build products for other developers, no matter how much they try the end result will end up appealing more to developers than the general population. This is illustrated by the weak show of Linux on the desktop, for years they've tried but not much has changed.

I'd like to think if we ever got a sizable enough install base on an open technology, the raw user base would sustain development through bug bounties and feature sponsoring. I still believe the most pure form of software development is someone paying you to fix their problems on a public codebase. They want it fixed, they pay to see it happen, and everyone benefits.

The problem, though, is that almost all consumer facing software doesn't see fundage through its user base - Photoshop, for example, is probably one of the most pirated pieces of software in history. It makes money through businesses buying business contracts for tens of thousands or millions of dollars to deploy it in the enterprise because they can't risk the gray space in buying one copy and sharing it around the office.

And businesses aren't going to be charitable to fund the development of their tools, even if they would be saving tons of money over paying out corporate contracts to the likes of Adobe.

> If you start with GPL stuff (and in my experience most of the really good open source software is GPL)

Much as I prefer the GPL to BSD licences, that sounds odd to me, unless you mean for pretty small values of most. Some best-in-class Free software more or less off the top of my head, that isn't GPL:

    apache (most everything, but lets stick with
      httpd and solr)
    nginx
    openssh
    freebsd
    openbsd
    openssl
    ghc
    go (golang)
    plan9
    PostgreSQL
    redis
    riak
    elasticsearch
    clang/llvm
    v8/chrome

If you add inn LGPL the list grows, perhaps notably to include erlang and to a certain extent haproxy (I didn't realize licensing had changed, the rationale may be interesting in this context[1]).

I guess there are a couple of ways to productize GPL software -- one is the RedHat/Canonical approach: use (rather large) resources on testing and creating a coherent whole, including branding (arguably RedHat does more work here than Canonical -- but Canonical certainly aren't the first to try to value-add on top of Debian). The other way is to sell an appliance of some kind. You could say the kindle and even Android fall under this category (as does haproxy, I guess).

> In such a scenario, any business that needed an off-the-shelf 3D editor or image editor but also developed in-house software (for example, Pixar) would find it in their interest to patch bugs and generally improve the software that they use, even though they don't sell it.

Absolutely. I think having a base on which to build is crucial for it to be even interesting to release certain software. The effort of releasing a usable version, needs to be as low as possible. Already having an open framework that one can simply "patch into" helps with that, I think. As does having a framework for contributing, and a community/project that will help with testing/merging etc.

[1] http://haproxy.1wt.eu/download/1.3/doc/LICENSE

The redhat model is really an outgrowth of the open source consulting model. If you look at smaller companies there are hundreds of projects that have spawned companies doing consulting, custom development and professional services.

Red Hat is unique mostly because they have been able to grow to this massive size. Part of this was initial capitalisation, part of it was right time, right place and mostly the fact that they have been able to add value to end customers more then their competitors.

XenSource had the problem of many other providers taking their output and repackaging it into a better product for customers (Amazon takes Xen, makes AWS. Other companies build their own virt products around it. If Xensource was the best virtualization platform and people had flocked for it, they would have been able to build a solid revenue from enterprise customers who required the guarantee of support for their mission critical workloads. Unfortunately there are lots of virtualization platforms around and I never found anything compelling in XenSource's Offering. VMware & HyperV have all the 3rd party support, oVirt / RHEV / OpenStack fill the cheap end.

I accept that with what the market wants Open Source companies of the size of Red Hat are rare, mostly because investors want larger returns, however this doesn't mean a open source company can't be profitable and grow to this size, it will just be significantly slower. Companies in the space typically will have a slow but solid growth pattern when firing on all cylinders, Not the 'hockey stick' growth curve many investors are looking for.

I've been wondering for a while if a company can work on purely open-source software and profit from the software itself, rather than only from the stuff that happens around it. The idea that I've been thinking over is a subscription for voting on features. That is, a user of the software will pay some monthly amount and get some number of votes for directing what new features will be implemented. Making sure that votes work properly and oft-requested features by people with low voting power still get implemented timely is one problem. Another is for companies that just want one bug fixed or feature added and aren't interested in long-term contribution. They could be easily outvoted by long-term contributors.

Has something like this ever been tried and how did it work or fail?

Unicon Inc, http://www.unicon.net/, partly does this. They cater to higher education with software like uPortal, CAS, and uMobile. All of these software products are open source, but they make money off of a support model like this. Universities that pay them for support can vote on features that they want worked on. I do not work for them, but my employer does use them for our student portal.

What a ridiculous comparison of apples and oranges and peaches and pears and plums and bananas. Amazon isn't even remotely in the same business as Red Hat. Microsoft is in the business of providing enterprise operating systems, but that's far from it's only line of business. Using Oracle as an example of how "selling support for open source software" isn't a repeatable business model ignores how Oracle is copying Red Hat's model, and in fact some of Red Hat's source code.

Is it true that Red Hat doesn't have the money to invest in improving RHEL? I don't know (although I suspect it's not true), but the article spends zero time on anything that would demonstrate that.

What prevents RedHat itself from transitioning to the open source platform model this article touts? They've put out OpenShift, which is a Heroku-style deployment clone, so they're at least making steps in this direction.

There are other RedHats already, SUSE and Canonical among them. So the article is flawed halfway through the title.

You could apply logic similarly to why there will never be another Microsoft.

If you look at it, Microsoft was a tremendous outlier. Not only was it the biggest licensed software giant, it was the only licensed software giant. By profit margin, until Apple and Google came along, nobody could compare. Cranking out CALs is literally printing money. And by revenues, the only companies which could compare were principally large integrated service providers: Oracle, IBM, SAP, PeopleSoft and Big N consulting firms' software divisions, producing custom software solutions selling brains by the bucketful (a notoriously difficult business to scale).

As wmf notes here, Red Hat's stated aim was to take the money out of the OS market (and put it in users' pockets), a task with which they were eminently successful -- they de-monetized Sun right out of existence, along with most of the proprietary UNIX market and much of Microsoft's server side.

The real value of free software is in enablement, which is why I see Google, Yahoo, Amazon, Facebook, and other companies built on the back of free software as its true legacy.

Of the companies Levine compares RH to, Microsoft is stumbling badly and Oracle has pretty much proven itself the place free software projects go to fork. VMWare is based on services to the free software (and proprietary) world, Amazon I've already defined as being build on top of free software.

This article is sorely nearsighted.

I thought I'd written about this elsewhere. Turns out I had:

http://np.reddit.com/r/dredmorbius/comments/1wsc0q/microsoft...

Oh golly, gee whizz, I guess we should just close our doors and shut down then. Peter Levine thinks you can't make money as an Open Source company... well, that sums it all up.

Wait a minute.

Fuck that.

I know Peter Levine is smart and well-respected and all that, but this is so much bullshit. For starters, all he really said, for using all those words, is "it's hard to build a business to the size of Oracle or Amazon or Microsoft, around a core of Open Source".

No. Fucking. Shit.

Really?

It's hard to build a business like Microsoft, Amazon or Oracle??? You're kidding me, right? I mean, it's not like any of those companies are atypical outliers in any way, right?

So, if we just take all our source, close it, and move to a proprietary business model, then we should have no problem building "the next Microsoft" right?

Also, am I wrong in thinking that Amazon is hardly even in the same business as Red Hat? What are they even doing in this comparison?

I dunno, color me biased (I am) but isn't Peter really just arguing that an "open source" company isn't going to generate the returns necessary, in the required timeframe, to justify investment by Andreesen-Horowitz? Because honestly, that's all I'm hearing. Nobody says you have to become "the next Microsoft" to be successful... well, nobody except Peter Levine, I guess.

Meanwhile, SugarCRM, Alfresco, Red Hat, Cloudera, BonitaSoft, JasperSoft, Pentaho and a whole laundry list of companies are making money "selling open source". Are any of them going to IPO? I don't know, but from where I'm sitting, that isn't the point.

All of that said, where I wholeheartedly agree with Peter, is the bit about adding a SaaS element with the underlying technology as a platform. At Fogbeam, I expect we'll have a traditional "support subscription" (ala Red Hat) model going for a long-time to come, but we are definitely starting to move in the direction of building purpose-specific / vertical aligned solution on top of our base stack, and delivering those as SaaS offerings. Personally, I see those as complementary strategies, and not mutually exclusive.

I don't really see why this may be a bad thing, the open source community continues to grow just fine and code contributions and stuff still happen. When I first read the title thought he'd be talking about how it's impossible to make money off opesource but that wasn't the case

Who cares about revenues? Of course the revenue is going to be higher at companies that sell software and hardware rather than support it. Costs are going to be higher too, for a company that has to write all of their software without help.

I started reading this article waiting jadedly for the inevitable call to move to "X as a service!!!!1oneone".

Sure enough, the article did not disappoint.

XAAS pretty much requires a cloud-based system, and public trust in the cloud is rapidly diminishing.

I just realized that there is another point that Peter totally skips over in this analysis... the value of community good-will, and the effect that has on the economics. For example, I'd bet money that a far larger percentage of Red Hat's customer base fundamentally like Red Hat, than the corresponding percentage of customers who like Oracle or Microsoft. Maybe I'm wrong, but I suspect that RH have lower customer acquisition costs and don't have to spend as much on advertising, since their product itself (in open-source form) is advertising for the company.

Joel Spolsky talked about building a free platform and living off the exhaust fumes here:

https://news.ycombinator.com/item?id=6462430

That's really the RedHat model. They just happened to get in early in an area that scaled well.

I agree with Levine's conclusions, but the reason Red Hat is successful in spite of "selling a slightly better road" is simple. (The car analogy really helps this.)

Unlike the other open source efforts Levine lists, Red Hat sells Free Software. Now, hang on just a second and hear this out: it's not that Red Hat is _evangelizing_ like the FSF does.

Red Hat has simply capitalized on the model that GPL Linux is better than _any_ proprietary server OS. I know I've never seen anything come close to it. Feel free to point out a server OS that competes successfully with Linux – gratis!

Red Hat can't charge for the razors but they make up for it by charging for blades: stuff your company _shouldn't_ be doing in-house. You can outsource it all to Red Hat: installation, maintenance, QA, everything. They're basically an early freemium model.

Levine is correct that the market potential has not exploded for Red Hat (yet) but he concludes that nobody else should try what they're doing. I think he doesn't understand exactly what Red Hat is doing.

Due to the unique pricing structure caused by Free Software, a "differentiated service on top of the platform" model would fail. For Red Hat, that would be something like selling proprietary system management software on top of Linux, and assuming it was sufficient.

If Red Hat were evangelizing Free Software they might be content with a sub-par product "because freedom!" Take laptops, for example: System76, ZaReason, and laclinux.com remain small. Their product does not represent a tier 1 product, even though they make a differentiated laptop on top of the same base of laptop ODMs.

Levine correctly concludes that Red Hat will not grow as rapidly as a Google or Amazon. It's not an apples-to-apples comparison. Instead, compare Red Hat to Microsoft, CA, or Oracle who compete directly in the "razors & blades" market of enterprise IT support. (Those companies do have other markets as well.)

Levine clearly understand's RedHat's business model: Most of their money comes from support contracts.

He also understands that selling support isn't as profitable as selling software, simply because the businesses scale differently. With software your costs are basically fixed (you sell the same thing many times), while with support your costs grow as the business grows (you need to hire more & more people).

Instead, compare Red Hat to Microsoft, CA, or Oracle who compete directly in the "razors & blades" market of enterprise IT support. (Those companies do have other markets as well.)

The numbers on those businesses make a good example. Microsoft makes almost all it's money from selling products, not support. Oracle has a very aggressive service division, and yet still makes more than 50% of it's money from selling software (and now hardware).

Most of their money comes from support contracts.

In my experience that's true of proprietary vendors as well. They just usually call it "maintenance" instead of "support".

That dichotomy is nowhere near as clear cut as you're making it seem. You think proprietary software vendors aren't also selling support? And don't need to hire more people as they grow? Bunk... I've worked for a few proprietary ISV's and from what I've seen, what they are doing is almost exactly the same thing as what Red Hat are doing. With Red Hat you pay for a "subscription" and with most other companies you pay for a license + "maintenance"... and the maintenance bit is, in essence, exactly what Red Hat are selling. The structure is just a bit different in that you don't, for example, pay for version upgrades in the RH model.

Nonetheless, even a proprietary vendor has their costs go up as they acquire more customers unless they refuse to support their software at all, in which case I doubt they are getting a lot of customers.

In my experience that's true of proprietary vendors as well. They just usually call it "maintenance" instead of "support".

Of course. This isn't an open source vs proprietary thing, and every company has to include a certain amount of support in the software they sell.

That dichotomy is nowhere near as clear cut as you're making it seem.

Agreed - I didn't mean to make is seem an "either/or" thing.

Nonetheless, even a proprietary vendor has their costs go up as they acquire more customers

Yes. But if your business is selling software your support costs (hopefully!) don't increase as quickly as your revenue. Eg, I don't know how many people Microsoft has as support staff, but I'll bet that number didn't increase linearly as revenue grew.

In the support business you will almost alway have a fixed number of people per license - or at least per customer.

The article isn't very well thought through. mongodb is doing very well just selling support

What about Cloudera?

Reply on news.ycombinator.com