Open core is not open source and don't trust someone trying to convice you otherwise

Oh my. I was outside painting my house for a few days, and when I return back online I discover that now everyone is having an opinion on the open core business model. Since some participants are still trying to promote it as a valid open source business model, let's see what everyone is saying and highlight any pitfalls being offered...

ladder in open and locked states

Before looking at all the links, let's just repeat what I said a week ago. While many have this week labeled the ongoing discussion as "the open core debate", using the word "debate" is rather questionable. The term "open source" was invented and well defined in 1998. Since then we have not had any debate over whether the definition should be changed, and there is no such debate today either. Open core is not open source as per the definition, and this is not up for debate.

Simon Phipps, formerly Chief Open Source Officer of Sun and current Director of Open Source Initiative, writes a ComputerWorld article Open Core Is Bad For You. Simon is putting open core high on the OSI agenda of dangers that the open source community and general public should be educated about:

In articulating the challenges facing open source after ten years of success, I asserted - as I usually do - that "open core" is one of the big challenges facing open source. This surprised some delegates.


Mårten Mickos then came back bravely defending the open core model (and still trying to qualify it as an open source model) in an article titled Open Source Needs A Little Closed Source To Win.

Mårten's points are familiar to me, since this is a discussion that of course was had many times already inside MySQL AB. Mårten is quite smart and good at argumenting his point of view, so I'd like here to highlight three of his arguments that look fine at first sight, but are in fact just attempts at making you swallow his closed source medicine.

First of all, it is worth repeating, that open source does not particularly need anything at all from Mårten. As Stephen O'Grady from Redmonk documents quite well, the most successful open source projects (like Linux and Apache) are not Venture Capital backed businesses to begin with, and even among companies, the most successful ones (Red Hat, Canonical) are not open core companies but so called "pure play" open source companies that have committed themselves to stay open source.

It is in my opinion great that investors are interested to put their money into open source enterprises, this can only help further boost open source. But for those that want to come and take over an open source project and make it closed source, there should be only one simple answer: Thanks, but we don't need your "help".

The second argument Mårten uses is the "we need to experiment with different business models" argument:

To have many companies that benefit massively in the open source space, I believe we have to practice many different business models. What works for Red Hat may not work for MySQL and what works for MySQL may not work for MuleSoft, and so on.

First of all, please don't try to associate yourself with Red Hat anymore. The suggestion that open core is just a variation of what Red Hat does, is false. Red Hat has committed to remaining open source and has innovated a business model that is compatible with open source. With MySQL you didn't do that, rather you created a closed source business out of it.

It is of course true that not all open source companies can just copy Red Hat's model. I personally believe there is still lot of innovation to be done in the area of open source business models. (And feel free to contact me if you'd like some consulting in this area, I have several unused ideas to share.) But open core is not such an experiment. Whether it turns out to be a good or bad model to make money (even this is still doubtful, as pointed out by O'Grady), the answer doesn't matter to someone interested in open source, since in any case it is not an open source business model, and no amount of experimenting will change that.

This indicates that open core companies operate within the major forces of free and open source software, and so it also indicates that the market is self-adjusting.

Mårten's final argument is that in any case, there is no need to worry about open core, since "the market" is self adjusting. This is of course true, but it is yet another devious argument to be aware of. (And it seems to be one of Mårten's favorites now: He also used it and the EU actually grasped this straw in the Oracle-MySQL decision.)

The fact that a market is self adjusting is not a get-out-of-jail-free card to do anything you want. For instance, to illustrate with an analogue from the domain of general ethics: Arguably our society is self-adjusting against various crimes like theft. We don't approve of stealing, and there are several measures against stealing, in particular laws and criminal punishments. Yet, from this it doesn't follow that stealing is only a crime if you get caught!

So similarly, if you put closed source modules into your open source product, and nobody notices, then you're still not open source. Like @Codepope (DJ Walker-Morgan, Editor in Chief of tweeted: "I like to treat the term open source like the term organic... You can't have "mostly organic" veg."

(Update: The above paragraphs were edited Sat Jul 3rd, to clarify that I do not intend to compare open core to stealing or imply that it is criminal. The illustration is that theft is against values of our society, whereas open core is against values of the open source community, but no further similarity was inferred.)

Ok, I think I made my point about Mårten's arguments. Clearly, he wants to associate himself with open source, still he doesn't want to give up on his closed source business model.

Matthew Aslett of the 451 Group and Stephen O'Grady of Redmonk then both chime in with analytical but laid back commentary. Both of them seem to stick to their analyst credo of mostly observing what happens and not having strong opinions either way. Aslett calls the whole debate "futile" since the best models will eventually win anyway.

Even so, both of them provide very good comments and new insights to the debate. Such as Aslett's descrption of Appcelerator's web service based strategy:

What is most attractive about this strategy is that it manages to provide additional value to paying subscribers without actually witholding any of the features or functionality of the core product from the community edition users.

This is at the heart of what this is all about, and thank you again to Matthew for this enlightening example. One thing that always bothered me with open core, that it is not at all difficult to come up with a strategy that is compatible with the open source definition, and therefore I see adopting open core as a sign of laziness and complete lack of commitment to open source. An important ingredient of countering the open core proponents is therefore to illustrate with examples of all the successful companies that are compatible with open source ideals.

O'Grady's article is similarly thoughtful and informative. It also suggests that advocacy against open core will be a futile attempt, and open source developers may as well just shut up and focus on coding.

This part I cannot agree with. Like Simon Phipps points out, open source (and free software / software freedom before it) was built on advocacy!

Even if I said that since 1998 we have not discussed changing the definition of open source, it doesn't mean that there haven't been attempts from proprietary software companies to ride on the open source momentum without really being open source. Microsoft had it's Shared Source initiative. Sun first used a license that didn't qualify as open source for its projects. There have been peculiar attempts like the SugarCRM license forcing you to forever advertise SugarCRM.

In each of these cases the open source community forcefully advocated against attempts to water down the open source brand and has been successful. Microsoft is no longer talking about the Shared Source initiative and have in fact in absolute numbers of lines of code contributed quite a lot to many real open source projects (of course the amount is still miniscule relative to Microsoft's size). Sun got the message and fully committed to real open source. And SugarCRM changed to standard GPL3 license to stay open source compatible (but SugarCRM is also open core, so not all modules are GPL3...).

Advocacy against open core will be just another step in the continuum against attempts to ride on the open source brand without being truly open source. This is in the interest of users and developers in the open source community. It is by the way also in the interest of those companies that have committed to open source. It is to their advantage that customers will prefer true open source products rather than the disguised closed source companies. Commitment to open source is their competitive weapon and now, just like before, it is in their interest to ensure the bar for qualifying as open source is not lowered to include closed source.

Giuseppe Maxia (not verified)

Fri, 2010-07-02 13:08

You seem to imply that MySQL is not open source.
So, please, tell me, which part of the MySQL server doesn't qualify as such?

Because, if MySQL is not open source, neither are the so called forks that are packaging 95% of its code with their own patches on a monthly basis.

(Posted by site administrator due to technical difficulties.)

Giuseppe, my friend. It seems your instinct to defend MySQL is stronger than the willingness to actually read what is being said...

First of all, at least the following parts of MySQL server apparently are not open source:
- The Windows installer
- The OS X installer
- The manual

...was there something more?

In any case, that is not even relevant to this discussion. The discussion is about the MySQL business model, and in particular MySQL Enteprise, which is the open core business model. The software that comes with MySQL Enterprise (The MySQL Enterprise Monitor and MySQL Query Analyzer) is closed source and I don't think you tried to deny that? This is also *all* of the software that is "in the package" for paying customers, since the actual server is identical to customers and free users. Hence, the MySQL business is about selling closed source software.

For completeness I'll briefly answer your second claim even if this is obvious: Forks of course cannot include this closed source software in their offerings.

Since you asked:

The windows installer is released under a license that does not require redistribution. AFAIK, no Windows user has asked for such code so far.
Why it was not redistributed? Because it is intimately tied with the internal tools that we use for packaging, and cleaning up the code would have been too much trouble. Since there is no requirement to release it, we have kept it in house. There is a next version in the making. If it turns out to be more modular than the current one, it might be released to the general public. No promises, though.

The Mac installer was checked in the repository by Lenz Grimmer in 2003.

bzr cat Build-tools/Do-pkg

It was removed in 2005, because since then we have been using Apple proprietary tools for the installation. Mac users are happy because the installer plays well. Also in this case, no user has complained that the code was not available.

As for the manual, this has nothing to do with the code. It has never been open source, since the beginning, but it can be redistributed freely as long as it goes with MySQL binaries. I am not particularly pleased with this choice, which was made long before I became an employee, but I understand its reasons and I support it, because I feel that the final users gets everything they need to use the server. And my feelings are the feelings of a satisfied user, former customer and partner before being a MySQL employee.

Thanks for the background. It is interesting to know about it for MySQL background, but I just want to reiterate that these are not the essential software that is relevant for the open core debate. The fundamental software for the open core debate is what is in MySQL Enterprise.

The "nobody has complained" part is kind of relevant to this blog post though.

Soon InnoDB hot backup will also be in the package of software for Enterprise customers.

I think you simplify this too much.

MySQL combines open source (MySQL RDBMS) with open core (query analyzer, hot backup, installers). I don't depend on the open core parts. I don't the installers and hot backup at work. I don't need QUAN at work (sorry, but it is a nice product).

I support open core as long as it is done via open interfaces. That allows the community to respond with things like Percona XtraBackup.

I think you simplify this too much.

I simplify to get the point across. If you want to debate, for example, whether MySQL AB was good or bad for open source, then my opinion, despite all this, is that it was on average very good for open source or that in any case the full picture is not a binary black and white. (For instance, since most people don't know what open core is or that MySQL was such a vendor, then in their minds MySQL's success simply was a sign that open source is great... which is a good image to spread, even if unfounded in that particular case.)

The right way to look at my arguments is in the context of open source business models, not so much open source software. Since the business is to sell closed source software, I think it is wrong for an open core vendor to self describe as an open source vendor.

Btw, it seems not everyone is convinced that the Hot Backup interface was made particularly open: (Ok, so apparently Heikki is contesting Domas' allegation.)

Hi Juanjo

When talking about open core we do not include business models where everything that is distributed to the customer is open source, and there is some proprietary web service or SaaS offering. Open core refers specifically to the practice where what is shipped to the customer includes closed source.

What you point to is also an interesting trend to follow, but it is important not to mix it with open core, otherwise the discussion gets convoluted. So we have to invent another name for it. My suggestion is "Proprietary SaaS offering".

this and this post from the 451 group for an overview of the different categories on business models.

About the bookAbout this siteAcademicAccordAmazonAppleBeginnersBooksBuildBotBusiness modelsbzrCassandraCloudcloud computingclsCommunitycommunityleadershipsummitConsistencycoodiaryCopyrightCreative CommonscssDatabasesdataminingDatastaxDevOpsDistributed ConsensusDrizzleDrupalEconomyelectronEthicsEurovisionFacebookFrosconFunnyGaleraGISgithubGnomeGovernanceHandlerSocketHigh AvailabilityimpressionistimpressjsInkscapeInternetJavaScriptjsonKDEKubuntuLicensingLinuxMaidanMaker cultureMariaDBmarkdownMEAN stackMepSQLMicrosoftMobileMongoDBMontyProgramMusicMySQLMySQL ClusterNerdsNodeNoSQLNyrkiöodbaOpen ContentOpen SourceOpenSQLCampOracleOSConPAMPParkinsonPatentsPerconaperformancePersonalPhilosophyPHPPiratesPlanetDrupalPoliticsPostgreSQLPresalespresentationsPress releasesProgrammingRed HatReplicationSeveralninesSillySkySQLSolonStartupsSunSybaseSymbiansysbenchtalksTechnicalTechnologyThe making ofTransactionsTungstenTwitterUbuntuvolcanoWeb2.0WikipediaWork from HomexmlYouTube