Comment by jillesvangurp | Hacker News Reader

Comment by jillesvangurp | Hacker News Reader

Parent Full thread

Redis ownership is decentralized as well. Redis the company owns the trademark and they were responsible for about one fifth of commits in recent years. The majority of contributions is external to them. The code base is BSD licensed and anyone is free to create a fork and continue development. Which looks like it is exactly what will happen now that Redis has decided that they don't need the input of the other 80% of commits from external contributors. Those contributors will inevitably shift their attention to one or several of the other forks. The Linux Foundation's Valkey fork looks like one of the main likely candidates for committers to rally behind. The biggest change will be the name change. The notion of the likes of Google, Amazon, Microsoft, etc. abandoning their Redis user base is unthinkable. They'll continue to offer that and they'll continue to lead the development of Redis. Redis the company will loose what little influence they had. I've never heard of anyone using hosted Redis directly from them.

In general, projects like Postgresql and other successful open source projects have in common that generally contributors are not coming from a single company. Community diversity is what gives open source projects resilience against corporate shenanigans. Mysql in some form or another will persevere as well. It has survived a lot of this stuff already and it's still there as an open source option.

I use it as a guide to select which things I use. I look for three things in open source tools and libraries that I use: 1) proper licensing (No agpl or shared source nonsense) 2) community diversity (no single companies that can change their mind), 3) active & recent development demonstrating the project is healthy.

Of course the tragedy of individuals like Simon Riggs passing away is that they are so important for the health of these projects. With postgresql, I'm confident that there are others that can step up. But still, he's been very important and it's important to recognize their amazing contributions. The OSS world is full of these type of hero developers and it's what makes using OSS so wonderful. With Redis, that person would be Antirez. And he stepped back from Redis the company some time ago. It will be interesting to see what he does post fork.

I’d guess Microsoft ends up replacing Redis with Garnet[0] in their stack at some point.

[0] https://www.microsoft.com/en-us/research/blog/introducing-ga...

What’s wrong with AGPL

AGPL is fine for open source projects. It isn’t really useful for a commercial closed-source or even an open-core codebase in some cases.

On the other hand, usually that’s the intention when a project selects AGPL. There’s usually a commercial license you can buy instead (see iText for example).

AGPL is absolutely fine for commercial closed-source projects. Don’t fall for anti-AGPL propaganda.

https://drewdevault.com/2020/07/27/Anti-AGPL-propaganda.html

It isn't fine because it will come up in due diligence as a risk, every single time. It's easy to say "it's fine", and in theory I agree with you. But it's a cost that doesn't make sense if you know what the auditors are going to say and expect. Auditors flag even more benign things than the AGPL, but to pretend it's not a thing is just whimsy.

I very much want drew's post to be correct, but I can't let confirmation bias cause me to consider this due diligence.

Are there any lawyers or legal cases that you know of that have proven this? or at least lawyers who have reviewed it and given a legal opinion?

It has a few clauses and language in there that generally scare corporate lawyers. There are two main groups of people advocating the use of this license.

1) companies that sell non oss commercial licenses for their AGPLed software that they own the copyright to and want you to buy those. A lot of those companies are now starting to prefer shared source type licenses.

2) open source advocates that don't like any commercial usage of their software and will actively want to prevent any form of intermingling of closed source and open source components like is common in many commercial projects. This is nominally to protect their freedom. But of course it has consequences in the context of commercial projects that don't want to opensource their proprietary stuff. Whether that is actually true or not for any particular use requires a bit of careful legal scrutiny.

Some places that do license audits (e.g. most banks, insurers, and other large companies that need to be alert to potential legal pitfals) would probably flag anything under this license. Three reasons for this: these licenses are only fine under very specific circumstances and certain combinations of licenses are not compatible. And finally of course these audits and lawyers are expensive. So, the easiest way to stay safe would be a blanket ban on anything with this license. Which is my general attitude towards this license.

Anyway, don't take your advice from random commenters (including me) on hacker news and consult a lawyer when in doubt. Yes that costs money. Alternatively, save some money and just steer clear of this mess.

> MySQL

What‘s the point of it, by the way? Why would one start a new project based on MySQL instead of postgres today?

MySQL has an actively maintained LSM-tree based storage engine like MyRocks, while Postgres doesn't have production-ready options in that regard.

Why is LSM useful for a RDB workload?

Log Structured Merge Trees are superior in write volume and scaling B+ trees higher. LSMs are part of the sauce of Cassandra and (I believe) DynamoDB for horizontal scaling.

RocksDB is a single node very efficient (better than Cassandra's in the 2.x/3.x releases, not sure about now) single node LSM implementation.

Because you know it better? Same reason as for most technical decisions tbh.

Besides that, there are still some points where the MySQL and/or the tooling around it simply performs better than Postgres. Anything to do with replication and big table migrations comes to mind.

Not an expert, but briefly looking into this, MySQL is

1. easier to configure and manage

2. faster for read-heavy workloads

3. has pluggable storage engine (though, if you care about this, then 1. likely doesn't matter anymore)

Just as a note, Postgres also has pluggable storage engine, using their tableam API.

It's much less emphasized compared to MySQL, I didn't even know about it until your comment, and search results are sparse.

MySQL has had multiple viable storage engines for most of its existence. MyISAM, InnoDB and more recently MyRocks.

You might be able to argue that Postgres' foreign data is a different form of pluggable storage engine. E.g. you can even use a CSV file as a backing store for a table with file_fdw.

https://www.postgresql.org/docs/current/ddl-foreign-data.htm...

https://dbaclass.com/article/how-to-access-csv-files-on-file...

Someone should do an experience report on adding the same storage system to both databases.

What you say looks powerful but Facebook added their MyRocks to MySQL.

Multi-master replication out of the box. Very useful, very occasionally.

Because I make money with WordPress.

I personally also prefer postgres but I have used mysql in the past as well on and off over the last 20+ years or so. Managed postgres was at some point slightly more expensive in Amazon for some reason. So that's a good reason. If all you need is some simple database, either is fine.

Because it may be good enough. In my case, because I already maintain an instance of MariaDB in production, and I may prefer to share this that maintain another thing

I would choose MySQL because I'm more familiar with it and it's good enough for what I need. PostgreSQL might be able to give me better benchmarks but it won't have any meaningful benefit for me the developer or for the user.

Benchmarks are not the reason people choose PostgreSQL over MySQL. I have never even bothered to benchmark the two for any project that could have used either.

Features, better ACID (at least historically) and maybe better standards compliance.

The big one for me is being able to run schema changes in transactions, which makes it easy to roll back a failed migration.

Also, the license. PostgreSQL is controlled by the open source community via a nonprofit organization.

True, but I would say governance rather than license.

Not being Oracle is a huge advantage ;)

I wouldn’t, only because the big O is behind MySQL.

Otherwise, why not? It’s basically a Chevy vs Ford decision.

Had Oracle stopped investing into MySQL, may be people would have moved on. But they didn't. Just like when people were worried about JAVA, instead we have 15 years of continuous improvement. And that is the same with MySQL. There are lots of features MySQL has as defaults, while postgres simply accept it is not something they want to deal with but leave it to extensions.

Reply on news.ycombinator.com