Thread: Feedback on blog post about Replication Feature decision and its impact
Hello everyone, I was about to blog about the announcement of adding replication to the core product but then thought I should ask for feedback here. I moved it to advocacy (from hackers). Any thoughts would be appreciated, thanks! --Dirk ---- Every software community has its peculiar challenges, and open source is no exception. This post discusses the relationship between a core open source product (PostgreSQL) and commercial offerings based on it (e.g. EnterpriseDB).<p> PostgreSQL is a relational database system (the most advanced open source database in its own words); to many it is known as the other open source database (next to MySQL). Unlike MySQL, PostgreSQL is not owned by anyone, it is a true community project. What's more, PostgreSQL is not based on the GPL (license) but rather on the more permissive BSD license, which lets companies distribute the database plus enhancements without having to contribute back their code.<p> The core system until now does not contain support for database replication. Naturally then, there are plenty of extensions, many of them open source, that add these features. However, if you need replication, it is rather cumbersome having to add some extension and maintain it separately from the core distribution. So the pressure has been mounting to add replication to the core product, and the core team of seven committers finally made that decision at their recent users conference. The <a href=" http://www.nabble.com/Core-team-statement-on-replication-in-PostgreSQL-td17537053.html">discussion of these new features</a> is refreshingly unpolitical and focussed on the task at hand, as to be expected of a mature open source community.<p> <a href="http://www.enterprisedb.com/">EnterpriseDB</a> is a well-funded database startup whose product builds on PostgreSQL. EnterpriseDB adds many "enterprise-readiness" features to the basic PostgreSQL product, including database replication, and much more. One might argue that it is not in the interest of EnterpriseDB to have replication added to PostgreSQL as it reduces the differentiation between the free community product and the more advanced commercial offering. Why pay for EnterpriseDB if you already get what you need from the free version? Won't adding replication to the core product reduce EnterpriseDBs sales? This tension seems only to get worse when you realize that EnterpriseDB employs several of the core developers of PostgreSQL, suggesting a direct conflict of interest when making decisions like whether to add replication or not.<p> So they finally made the decision to add replication, and it gives me the opportunity to discuss what I believe are misunderstandings about the open source business.<p> <h3>Won't EnterpriseDB loose sales once replication is added to the core PostgreSQL product?</h3> I think the opposite will be the case. Officially, EnterpriseDB wants to be a cheaper Oracle, but in the open source arena, its main competitor is MySQL. EnterpriseDB the commercial offering is competing with MySQL the commercial offering, and not with the free community version of PostgreSQL. It is in EnterpriseDB's interest to have a free PostgreSQL version installed and used in as many IT departments as possible, because it is <a href="http://www.riehle.org/2008/04/30/sdn-is-open-source-competing-unfairly/">the first (and important) step to a later sale</a>, as I have discussed elsewhere. Enhancing the free product achieves exactly this.<p> <h3>Won't a reduced differentiation between EnterpriseDB and the core product reduce their addressable market?</h3> I'm pretty sure it doesn't. The addressable market size doesn't go down. That's because EnterpriseDB is not only selling additional features, but more importantly to many applications and customers, it is selling "operational comfort". Operational comfort means that EnterpriseDB is offering its throat (to choke) to customers should something go wrong. For money obviously; this is a core part of their business. Once a database system becomes mission-critical, few companies will want to go without paying for support. What the reduced differentiation does, however, is to increase the possible competition around selling such operational comfort. Other companies may more easily enter this market and compete with EnterpriseDB. However, as I have argued elsewhere, <a href="http://www.riehle.org/computer-science/research/2007/computer-2007.html">by employing core developers, EnterpriseDB is well positioned to make a believable case that it is the go-to provider of operational comfort</a>.<p> <h3>There is only one license, the GPL, and everyone should be using it.</h3> PostgreSQL is a good example of a community open source project that does not use GPL and still is flourishing well. Whatever the ideological background of this statement, the belief seems to be that people should be forced to contribute back to a project rather than do so of their own choosing. That's hardly a notion that increases freedom. More importantly, the rationale behind it makes little sense to me. In general, firms and individual contributors alike are motivated to contribute back (non-differentiating) code to reduce their maintenance overhead. If they don't, they'll only create more non-differentiating work for themselves as they are trying to catch up with the evolving codebase. What's more, every possible proprietary extension faces the problem of possibly being contributed by someone else, if only there is enough demand for it. If you wait too long to make your contribution, someone else will do it, and you just created another maintenance and migration problem for yourself. Which is exactly what we see happening with the replication feature in PostgreSQL. Pressure had been mounting, and now it will be included, to everyone's benefit.<p> -- Into novel software paradigms, tools, processes? Then submit a short paper to Onward! 2008 by July 2nd! See http://www.oopsla.org/oopsla2008/cfp/cfp-onward.html -- Phone: + 1 (650) 215 3459, Web: http://www.riehle.org
Dirk, > <a href="http://www.enterprisedb.com/">EnterpriseDB</a> is a well-funded > database startup whose product builds on PostgreSQL. EnterpriseDB adds > many "enterprise-readiness" features to the basic PostgreSQL product, > including database replication, and much more. The replication-in-core vs. not-in-core has absolutely nothing to do with EnterpriseDB either way. I think you'd be doing a disservice to your readers by implying that it does. Or with the GPL. If you want to blog about these things, maybe break them up into seperate posts? -- Josh Berkus PostgreSQL @ Sun San Francisco
On Saturday 31 May 2008 14:48:26 Josh Berkus wrote: > Dirk, > > > <a href="http://www.enterprisedb.com/">EnterpriseDB</a> is a well-funded > > database startup whose product builds on PostgreSQL. EnterpriseDB adds > > many "enterprise-readiness" features to the basic PostgreSQL product, > > including database replication, and much more. > > The replication-in-core vs. not-in-core has absolutely nothing to do with > EnterpriseDB either way. I think you'd be doing a disservice to your > readers by implying that it does. Or with the GPL. If you want to blog > about these things, maybe break them up into seperate posts? > Adding any feature into core Postgres can have a ripple effect on EnterpriseDB, so I don't see any reason not to discuss these issues. I'd be more concern about the factual errors in the post (things like "core team of seven committers") than exactly where the OP thinks the discussion should go. -- Robert Treat Build A Brighter LAMP :: Linux Apache {middleware} PostgreSQL
Dirk Riehle wrote: > <h3>Won't EnterpriseDB loose sales once replication is added to the core > PostgreSQL product?</h3> > > <h3>Won't a reduced differentiation between EnterpriseDB and the core > product reduce their addressable market?</h3> > To these two I would mention that although the developers are planning to add basic replication with a goal of easy config, this won't take out the need for a more complex replication options that EDB offers -- Shane Ambler pgSQL (at) Sheeky (dot) Biz Get Sheeky @ http://Sheeky.Biz
Thanks for the feedback. I was a bit worried the post would be too bland, but I think I'm seeing now where I wanted to go in the first place, which is to discuss perceptions around conflict of interest. I sometimes hear stuff like "in order to take over an open source project you need to hire the committers". Ignoring the insult to the integrity of the committers, I think this is also based on a wrong idea of conflict of interest. One typical perceived conflict of interest is that commercial companies may use committers to keep commercially relevant features out of the free community product in order to better facilitate an upsell. One might have argued that replication in PostgreSQL is a case in point. So in my blog post I argue that enhancing the core product is actually in the interest of commercial offerings because "the enemy" is not the free community edition but rather alternative products like Oracle or MySQL. And making the free community product stronger beefs up the sales process, because a corner stone of open source based sales processes is to get free versions into potential customer companies. Now, I'm sure I'm a bit naive about this, however, the core argument I make above seems right to me. It would be interesting see where actual conflicts of interest happen and the last defense for the community is actually the integrity of the committers, and not some economic reasoning. Cheers, Dirk Josh Berkus wrote: > Dirk, > > >> <a href="http://www.enterprisedb.com/">EnterpriseDB</a> is a well-funded >> database startup whose product builds on PostgreSQL. EnterpriseDB adds >> many "enterprise-readiness" features to the basic PostgreSQL product, >> including database replication, and much more. >> > > The replication-in-core vs. not-in-core has absolutely nothing to do with > EnterpriseDB either way. I think you'd be doing a disservice to your readers > by implying that it does. Or with the GPL. If you want to blog about these > things, maybe break them up into seperate posts? > > -- Into novel software paradigms, tools, processes? Then submit a short paper to Onward! 2008 by July 2nd! See http://www.oopsla.org/oopsla2008/cfp/cfp-onward.html -- Phone: + 1 (650) 215 3459, Web: http://www.riehle.org
Re: Feedback on blog post about Replication Feature decision and its impact
From
"Jonah H. Harris"
Date:
On Fri, May 30, 2008 at 6:21 AM, Dirk Riehle <dirk@riehle.org> wrote: > <a href="http://www.enterprisedb.com/">EnterpriseDB</a> is a well-funded > database startup whose product builds on PostgreSQL. EnterpriseDB adds many > "enterprise-readiness" features to the basic PostgreSQL product, including > database replication, and much more. One might argue that it is not in the > interest of EnterpriseDB to have replication added to PostgreSQL as it > reduces the differentiation between the free community product and the more > advanced commercial offering. Why pay for EnterpriseDB if you already get > what you need from the free version? Won't adding replication to the core > product reduce EnterpriseDBs sales? This tension seems only to get worse > when you realize that EnterpriseDB employs several of the core developers of > PostgreSQL, suggesting a direct conflict of interest when making decisions > like whether to add replication or not.<p> Having worked for EnterpriseDB for about three years now, and understanding that there are several facets of our business model, I can say that adding these types of features wouldn't hurt us at all. In fact, they would make some of our work easier, and would be welcomed. -- Jonah H. Harris, Sr. Software Architect | phone: 732.331.1324 EnterpriseDB Corporation | fax: 732.331.1301 499 Thornall Street, 2nd Floor | jonah.harris@enterprisedb.com Edison, NJ 08837 | http://www.enterprisedb.com/