Is there any chance of us getting something like this going to support our Debian development efforts? What do you think of Hans' ideas? ---------- Forwarded Message ---------- Subject: [reiserfs-list] Viability of Open Source Business Models Tried Once Again:-) Date: Tue, 04 Dec 2001 19:05:11 +0300 From: Hans Reiser <reiser@namesys.com> To: reiserfs-list@namesys.com I feel better about it not being my error that the document was unattached from my email.:-) This is being translated into Russian for publishing, and if you would like to be the first english language publisher let me know. Best, Hans ------------------------------------------------------- -- http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark http://www.coker.com.au/postal/ Postal SMTP/POP benchmark http://www.coker.com.au/projects.html Projects I am working on http://www.coker.com.au/~russell/ My home pageTitle: The Viability of Free Software Business Models
Making your software free might DECREASE your service contracts revenue.
The free software movement is quick to cite service contracts as the best way to make money from selling free software. There is a problem with that.
Companies are quicker to spend money on service contracts for proprietary software.
The typical Linux software using company will have service contracts on its Sun workstations including the software, and a hardware only service contract on its Linux workstatons. It will have service contracts on Network Appliance fileservers, and, well, there is exactly one company (Lycos-Europe) with a large service contract on its ReiserFS fileservers.
If a Linux using company does have service contracts on its Linux workstations/appliances, most of the time that money goes to some intermediary company which doesn't give any of the money to the authors of the software being serviced.
This is not rational economic behavior. The authors of the software will indeed give much more expert advice than anyone else. They will probably give it cheaper also. It makes no economic sense whatsoever to be slower to buy a service contract on free software than on proprietary software if you use the free software.
It is very real behavior though, and it is killing off many free software companies. Sourceforge went proprietary because their sales people couldn't sell service contracts for free software, and making the premiere version of the software proprietary made the service contract an easier sale. GFS, Bitkeeper, how many pieces of software might have been free if the service revenues had been there?
This is deeply endangering the free software industry's fiscal viability.
Why are they reluctant to pay for service on something that is free?
I get several emails a day from users who want me to tell them something in our FAQ, or to diagnose that they didn't really recompile the kernel module like they thought they did, etc. I am happy to do this, but I think I should get paid for it quite honestly.
We have a support page which has to have the lowest fee in the industry, $25 a question, and we'll answer any Linux related question (though most of the ones we get are ReiserFS related quite naturally.) This support webpage gets used once every several days.
I refer the emailers with user error problems to the support page. (Bug reports not in the FAQ we handle for free, though the proprietary guys laugh and tell me that they charge for bug reports also.) Maybe 40% of those emailers with user errors send me a flame back, angry that I want money to answer a simple question that I know the answer to. They wouldn't send an email like this to their medical doctor expecting a free diagnosis.
So we have a real problem in our industry. Economically irrational behavior by users is killing our industry's revenues, and it demolished the business plans that all of those Linux VCs funded once and swore to never fund another one looking like it again.
Where do we go from here?
The advantage that Gnu/Linux has is passion. We need some focused passion. If it is not logical, then we need to get going with the psychological! If you believe in the value of the free software movement, then you need go to the next IT purchases meeting at your company or government lab, and ask if there is a service contract from the developers on all of the significantly used free software. It is that simple.
We need to create a cultural norm in which it is considered proper behavior to buy service from free software authors with the same frequency one buys it from proprietary software authors. The way to do it is to acknowledge we have a problem, and start telling people being irrational doesn't just leave you losing the cost of your time, it leaves you hurting a movement you care about.
There are other ways also, but I don't find them as attractive.
My guess is that most of the reason the proprietary vendors sell more service contracts is that users are more inclined to buy service when it is one more line on a contract than when it is a new contract.
An economist might be tempted to call this a transaction cost, and think rational behavior is occurring, but beware that temptation! This marketing phenomenom holds true even for customers where the cost of service would be hundreds of thousands of dollars. Rational economic behavior for the customer company as a whole it is not.
We could for instance modify the license to require users to purchase zero or more dollars worth of service from the software authors (including issuing a purchase order or supply a credit card even if the order is for zero dollars). I think that some of those vendors whose free software business plan seriously depends on service revenue should consider experimenting with this. Alternatively, we should just not depend on service revenue for the support of free software businesses.
Perhaps one can better make substantial amounts of money from service if one's product is complicated enough. MySQL seems to be making that approach work. Namesys plans to add database and hypertext functionality to ReiserFS. It is possible that someday our filesystem will offer sufficient functionality that having a Namesys consultant will be worthwhile for the average fortune 500 company. In that case the whole revenue model could snowball, with each consultant creating more features, each new feature creating more customers, and the end result could eventually be a company the size of Anderson consulting advising companies on their information bases. While there is nothing unique to open source about this, it is possible that open source could allow a faster initial user adoption rate.
It is interesting that free software service revenue at this time seems to encourage economic concentration. This is not very encouraging if it holds true for the long term, and we end up with proprietary software locking out competitors from the source code, and free software vendors either not giving any of the revenue to developers (most small vendors) or being highly concentrated large vendors who favor funding software internally developed and controlled. The only thing worse than a state controlled economy is a monopoly controlled economy.
Keep in mind that this is being written from a perspective in the depths of the 2001 tech bust in which Namesys is managing "only" a 100% annual growth rate due to the severity of the economy. While ReiserFS usage is increasing by one or two orders of magnitude, all non-governmental sources that were on the verge of increasing revenue by an order of magnitude disappeared as those customers entered into layoffs.
That does imply though that one must segment one's market. One way to segment the market is according to whether the software is integrated with proprietary software. Our licensing policy says that if you make money from charging for usage licenses, then we must make money too. This allows Namesys to sell to fileserver vendors and proprietary software vendors who want to add proprietary improvements to ReiserFS. All sales that Namesys has made in this area have resulted from market sampling being very effective. It generates far more revenue at this time than service does.
Sales to proprietary OS vendors have not happened yet, but once the economy recovers this could change, and interest has been expressed. It takes just one sale of that size to make 8 years of work on a high risk technological gamble pay off financially.
We could make proprietary optional plugins, with the core filesystem remaining GPL'd. An issue though is how to prevent the poor from being locked out of access to software. Software fees are usually set to what an average middle-aged man spending his boss's money at a US corporation would be willing to pay. Currently the only reasonable option for the poor for most software is to pirate their software. (Pirating can also serve as a form of anti-trust vigilante action when the judicial system is as corrupted as it is in the US.) Making license fees proportional to storage hardware cost is the best way to handle this.
This still leaves us with the economic inefficiency that occurs when a user needs a spreadsheet three times a year, and he would like to use the best one made, but the best one is priced appropriately for those who use it several times a week. I call this the focused taxation inefficiency.
Most products are sold for more than their marginal cost because most real markets are only semi-competitive. Doing this distorts the economy. Software takes this distortion to its greatest extreme, and best illuminates a longstanding inefficiency in real world semi-free market economics.
I have an unproven hypothesis to put before you: The amount of economic distortion is a function of not just the total dollar amount traded and exceeding the marginal cost, but is a function also of the ratio of price to marginal cost, and high ratios are "typically" more distorting than low ratios, even if the total number of dollars is made equal. Persons interested in discussing proving this hypothesis are encouraged to email me at reiser@namesys.com. If the hypothesis is true, then can we choose to defocus the disincentive?
Can we avoid the serious disadvantages of a state controlled, or monopoly controlled, economy, and avoid the focused taxation inefficiency at the same time? How can we lockout neither poor customers nor poor startup OS component suppliers?
I'd like you to consider the concept of open sales. In an open sale, users pay a fixed percentage of hardware costs as a fee for using a software pool. Users may use all software in the pool. The users are presented with a random sample of those developers who contributed to the pool, and asked to allocate their fee among the sample in proportion to their perceived utility.
Open source is a niche we at Namesys enjoy inhabiting. Its production is not motivated mostly by money. It is a gentleman's hobby for most, and a profession for some who aren't in it for the money. It is likely to remain a niche. As long as that is understood, a good time can be had by all us mutant hairless monkeys. It is amusing to give away for free a filesystem faster than what a company worth billions is able to write despite all their money. Life should be enjoyed.