<antirez>

antirez 2069 days ago. 134410 views.
Human beings have a strong tendency to put new facts into pre-existing categories. This is useful to mentally and culturally classify similar events under the same logical umbrella, so when two days ago I clarified that the Redis core was still released under the vanilla BSD license, and only certain Redis modules developed by Redis Labs were going to change license, from AGPL to a different non open source license, people said “Ah! Ok you are going open core”.

The simplification this time does not work if it is in your interest to capture the truth of what is happening here. An open core technology requires two things. One is that the system is modular, and the other is that parts of such system are made proprietary in order to create a product around an otherwise free software. For example providing a single node of a database into the open source, and then having the clustering logic and mechanism implemented in a different non-free layer, is an open core technology. Similarly is open core if I write a relational database with a modular storage system, but the only storage that is able to provide strong guarantees is non free. In an open core business model around an open source system it is *fundamental* that you take something useful out of the free software part.

Now for some time Redis is a modular system. You can use Redis modules in order to write many things, including new distributed systems using the recently introduced cluster message bus API, or new data types that look native. However the reason to make Redis modular was not to remove something useful from the system and put a price tag on it. For instance one of the new data structures in Redis 5, the streams, are part of the core and are released under the BSD license. Streams were implemented when Redis was already a modular system.

Redis modules started from a different observation. As a premise I should say that I’m a very conservative person about software development. I believe that Redis should be focused to address just things that, when operated with in-memory data structures, offer strong advantages over other ways to do the same thing. I don’t want Redis to do much more than it does, or to employ every possible consistency tradeoff. I want Redis to be Redis, that is, this general tool that the developer can use in different ways to solve certain problems.

However at Redis Labs we observed multiple times that it’s a bit a shame that Redis cannot solve certain specific problems. For instance what about if Redis was a serious full text search engine? Also well, developers want so much JSON, what about having an API to talk directly JSON? And given that in-memory graphs if represented wisely can be so fast, what about having graph database abilities, with a rich query language? Redis Labs customers often asked directly for such things. And actually, such features could be cool, but it’s not Redis, I’m not interested, and the open source side of Redis does not have the development force to keep all this things going btw. And this is a major advantage both for Redis and for Redis Labs: it is relatively cheap to pay just me and a few more OSS development time internally, while allocating the rest of the resources to development of things that are useful for the Redis Labs business, like making sure the Redis enterprise SaaS and products are good. There is anyway a great deal of contributions arriving from the community. And I also keep saying “no” to all this fancy ideas that would keep Redis in other areas… which is also a problem.

Still to have such things similar to Redis but outside the Redis scope, would be cool, because you know, while it’s not Redis mission, people may very well use a fast inverted index with full text search capabilities that you can feed in real time, while it serves a very good amount of queries per core at the same time. This is what Redis Labs is doing, it’s using the same Redis technology and approach to do more than what Redis wanted to do. Not just in the functionality areas, but also in other areas like consistency models. I’m very opinionated about certain things, and I think that, for instance, CRDTs while super cool in certain use cases, where not the right thing for Redis, to retain the same memory footprint, performance, simplicity, even at the cost of having a weaker consistency model. So Redis Labs, together with a top researcher in the area, did it (and this is a proprietary product without any source available). I can see how such feature can be tremendously useful for certain operations, but Redis was not there to solve everything, and Redis Labs did it.

This is not open core. Redis Labs is doing things that you would never see from me: for bandwidth, and because I believe that not all the softwares must eventually become huge. 
So I think that calling this model “open core” is misleading, nothing is removed from the Redis table, just new things are explored, while trying to follow the “Redis way” in other areas otherwise not touched by the Redis project.
🚀 Dear reader, the first six chapters of my AI sci-fi novel, WOHPE, are now available as a free eBook. Click here to get it.
blog comments powered by Disqus
: