The race to replace Redis [LWN.net]

The race to replace Redis

Posted Mar 30, 2024 11:25 UTC (Sat) by gmgod (guest, #143864)
Parent article: The race to replace Redis

Like others, I really like this article. It's exhaustive and laid out well.

Regarding the content and the situation, GAFAM has been fuding the hell out of developers to convince them BSD/MIT is the only way... Now they have their ways for minute or no contribution at all (what's missing from this article are the kind of commits that were contributed to "refactor core engine to yield 30% more perf on average" is the not the same as "add a hook to facilitate AWS integration").

I still remember ElasticSearch's lead going mental (as in "really upset and distressed") with how their open-ass licensing was supposed to bring in more contributions and yet only allowed Amazon to get fatter by selling their OSS. It looks like the same story is bound to happen again and again. If you are not happy with anyone else profiting from your software, don't use a license that allows it. If you don't want your lawn to be used by randos, don't open your gate and put a sign reading "public park".

So now the situation has nearly gone full cycle... Proprietary -> hard copyleft -> soft copyleft -> permissive -> ...

I'm sure in 5 years, the time it will still take for consulting to pick it up, the whole "what are you doing, BSD/MIT is the only way if you want contributions" will be transformed into "if you want to make money with it, go proprietary or choose a strong copyleft license so that you don't shoot yourself in the foot by providing downstream with all your patches and not requiring they do the same".

to post comments

The race to replace Redis

Posted Mar 30, 2024 13:21 UTC (Sat) by pizza (subscriber, #46) [Link] (6 responses)

> I still remember ElasticSearch's lead going mental (as in "really upset and distressed") with how their open-ass licensing was supposed to bring in more contributions and yet only allowed Amazon to get fatter by selling their OSS.

Amazon isn't "selling [ElasticSearch's] OSS" -- It is "selling a service built using ElasticSearch's OSS". That's a _crucial_ difference.

At the end of the day, the _software_ is in of itself irrelevant; what matters is the _services_ the software can be used to provide. When it comes to providing services, it's pretty much impossible to compete with the giant players' scale [1]. Especially when those giant players are (more often than not) deriving the majority of their income from datamining users of those services, an approach that is only feasible once you get to a certain size.

[1] Unlike software that is "one-and-done" with a marginal reproduction cost of effectively zero, providing services has an incremental per-use[r] cost (bandwidth, compute, storage; plus the ongoing infrastructure/management cost of robustly scaling).

The race to replace Redis

Posted Mar 30, 2024 18:42 UTC (Sat) by gmgod (guest, #143864) [Link] (1 responses)

I really don't see how any of this is in reply to what I said, on spite of the citations...

BSD/MIT and other permissive licenses allow anyone to do almost anything with the software and most of the recent (and more ancient) license changes we've seen all come down to "oh shit, big companies make big buck based on what we do without contributing back enough to ensure we are sustainable". Whatever added value those company may or may not provide is irrelevant.

The software is not irrelevant when it's part of the service. Try to repeat yourself gas, or breaks, are irrelevant when buying a car and see how far it goes.

Those people just realised it's hard to build a business model where all the work you do can directly be integrated downstream. Downstream thus becomes strictly superior the day they provide anything at all. That's not a good situation to be in when you try to earn a living for that work.

The race to replace Redis

Posted Mar 30, 2024 19:03 UTC (Sat) by pizza (subscriber, #46) [Link]

> The software is not irrelevant when it's part of the service. Try to repeat yourself gas, or breaks, are irrelevant when buying a car and see how far it goes.

(Most) Folks don't care about a "car" in of itself, they care about the "transportation service" it provides/enables. Gas or even brakes aren't a given; transportation services can utilzie altenative fuels (eg electricity0 or eschew friction braking in favor of alternative methods (eg regenerative, reverse thrust, etc).

Similarly, folks don't buy or download "software" for its own sake, they do so because it enables them to achieve certain things.

The race to replace Redis

Posted Apr 1, 2024 16:39 UTC (Mon) by immibis (subscriber, #105511) [Link] (1 responses)

The intention of the GPL is clearly that Amazon can do this, but they have to publish their modified and extended version of the software. AGPL fixes a loophole where they don't have to publish at all. SSPL tries to fix a loophole where they have to publish their modifications but not their extensions.

The race to replace Redis

Posted Apr 6, 2024 16:31 UTC (Sat) by Wol (subscriber, #4433) [Link]

> The intention of the GPL is clearly that Amazon can do this, but they have to publish their modified and extended version of the software.

WRONG WRONG WRONG.

Okay, the GPL was written for the days when SAAS was not a thing, but the intent of the GPL is quite clearly you have NO OBLIGATION WHATSOEVER to share, but sharing is an "all or nothing" proposition.

Cheers,
Wol

The race to replace Redis

Posted Apr 5, 2024 16:36 UTC (Fri) by nim-nim (subscriber, #34454) [Link] (1 responses)

> When it comes to providing services, it's pretty much impossible to compete with the giant players' scale

Actually that very much depends on the amount of margins those giant players want (and think they can afford) to charge. Even on-premises is competitive when the alternative is something someone else slapped a 80% services margin on. Never forget that the giant players are not so numerous and they would be quite happy to constitute an oligopoly for everything other players could not run themselves.

Going proprietary does not protect an ISV it ensures all the other players will invest in something else and gradually squeeze it out of the market (but the money is good VC-side while the squeezing is going on).

Long-term viability seems something open enough your product is accepted, and innovative enough part of the market prefers paying you for early access and a say in the product evolution. And there will be some freeloaders but those freeloaders will also keep cloud giants from making too much profit at your expense. One way or the other the level of margins an Oracle or Microsoft enjoyed before the internet seems out of reach now.

The race to replace Redis

Posted Apr 5, 2024 17:11 UTC (Fri) by nim-nim (subscriber, #34454) [Link]

Also consider this. None of the “traditional” ISVs have made their money on software dev for a long looong time. For a couple of persons that worked on code there was an army of people doing QA sales ticketing and some form of service deploying software on premises (and if you were too small to deploy yourself you’d better partner with some consulting giant that would deploy things for your customers or you would miss out most sales).

Cloud services are not so different you still have a smallish core of people developing things and a mass of people making money out of it by deploying code operationally. I’m not sure the ratio of core people/other people has changed so much.