Redis and the Cost of Ambition

Thank you for the thoughtful reply, I'm grateful you took the time to write it and share it. I think it's valuable to have the commercial way of thinking presented alongside to give this post the additional context - that Redis doesn't exist in a "pure software" vacuum, that OSS gremlins like me aren't their target audience, and that decision-makers often have different criteria than a systems eng might have. I feel like your comment answers the business perspective "why" (while the post is mostly describing consequences of this).

One thing I will mention, and perhaps this is exactly your intention, but the comment and argument sorta pre-supposes that money is everyone's goal in this. This is, to me, rather a specific ambition -- the ambition to make a shitload of money off of Redis in this case. I don't know that antirez ever cared about it, based on his writing I don't read him as someone who cared about money much.

Perhaps the finest example of this mentality is the SQLite project. They're not going to talk about millions in revenue or billion-dollar-sales. They are focused on delivering something quietly, which is well-defined. SQLite did not lose sight of why it became the most popular embedded database. Nor did Postgres, for that matter (on the "big" database side). That is just as valid a universe to draw counter-examples from as the money/commercial-interests side.

For Redis the "we already have AWS/GCP, let's just use whatever their version of it is called" is a valid extension of your horizontal expansion point. The more "IBM" path seems to me to pick the cloud provider and use their Redis - which happens to be Valkey these days.

That people choose shittier options is, I agree, not debatable - it's just a fact. Redis' expansion into that kind of "mode" is an instance of precisely the ambition/drift that I wanted to highlight.

I've worked at Redis Labs, so you could say I was almost literally the devil's (developer) advocate.

I can also speak with integrity because I have zero financial ties or motivations any longer to be anything but truthful.

When I left RL I decided to not exercise the options I vested, so not only I have no financial ties, but also this should give a hint of what I think of RL's future. Also I have the luck of living somewhere far away from the silicon valley, so I can speak my mind without fear of burning one bridge too many.

The Redis landing page to me looks like a TDM-oriented site. And the "real-time context engine for AI" and AI focus feels correct for that target user.

One reason why people are annoyed at this is because redis.com used to be the TDM site, while redis.io was for developers. After RL got its hands on redis.io (used to be owned by antirez), they tainted it to the point that you couldn't even find a link to download Redis anymore, and now both URLs send you to the TDM site (and you still can't easily find a link to download Redis, I challenge people to try).

My take is that the core problem is that RL has always hired chumps for marketing leadership. Chump CMOs and chump VPs that would come, burn as much goodwill they could in the shortest amount of time, and then leave 6 months later for 'the next adventure'. I never understood how the company could fail so reliably at hiring marketing leadership. In the case of the website I can promise you that somebody in marketing noticed that redis.io was getting a lot more traffic than redis.com, and so they did the thing that any chump would do: ruin redis.io in the hopes of getting that one extra click through (om nom nom all those delicious clicks on 'try free' from people that can't find the download link). Just another example of burning goodwill for minimal, meaningless, short term gains.

Then once the deal is landed, you have a cadre of add-on functionality that you're probably just average at at best, but its easier for procurement (the department that handles software purchasing in a business) to upgrade an existing closed deal than to engage in a new one. So you can sell mediocre stuff.

Definitely. It also helps sometimes to differentiate yourself from your competitors (at least checkbox-wise), especially when it's impossible for you to compete on price. AWS had the ability to offer ElastiCache packages that included (heavy, some would say anti-competitive) cloud discounts, and then there was the worst competitor of them all: open source Redis. Can't compete on price if the other offering is free :^)

Maybe, maybe not. Valkey is doing far better than the average fork, that's for sure

Valkey also has orders of magnitude more money being spent on it that most forks (RL's old Head of DevRel now works at AWS on Valkey, just as an example). And in fact this is the part of the original post that I find the most disappointing.

Valkey's existence and adoption is the wider market's final verdict on this dynamic. Rather than chase features and bullet points, Valkey has invested in the un-glamorous work of improving multi-threaded performance, memory efficiency, cluster reliability and throughput. Valkey's performance benchmarks are impressive and aimed squarely at the 80% of Redis users who just want the same features Redis shipped with back in 2011. There's no need for a new array type in Valkey's world.

First of all: Redis already has been multithreaded for a long time. Still single-thread control plane, but MT for I/O. Valkey's work in that sense is not all that novel as the author thinks it is.

Second: Valkey is blatantly an operation to make sure that AWS (&co) can continue selling Redis without having to pay a dime to anybody. That's the one and only goal, every other argument is just a tool that marketing will wield to convince you to let them keep doing what they want. And I promise you that if they decide that it's commercially advantageous to break compatibility with Redis, they will do so. I would actually be willing to place a moderate bet on the fact that they already have done so (on both sides), I just haven't paid close enough attention to know for sure.

If you want the real Redis fork that does un-glamorous work and that just wants to keep things simple, that's redict.

That being said, in my opinion the time of Redis is coming to an end. This new weird commercial landscape around it makes each fork stunted one way or another. Even redict, which is the most virtuous of the forks, has the problem that there is no genuine interest in pushing Redis forward the same way that it was for the original upstream project back when antirez was working on it as dictator. Not to say that it's bad for software to be "done", but personally I think that there are still some undiscovered cool things that Redis could be made do, and I doubt that the commercial forks will provide the right ecosystem for that to happen, but I might be underestimating the value of Arrays and their applications wrt AI, I'll try to keep an ear out and an open mind about it.

Thinking about it all, it's incredible how badly RL fumbled it all, good thing it's been long enough that I don't get angry about it anymore :^)