[Raion]

This is an open discussion. If you think I'm retarded, you are free to say so! This is mostly to answer some questions I get every so often casually:

1. Why do you like postgres? What's wrong with MySQL? or Mongo, CouchDB, Couchstore (insert tech here)?

Ok, so there's three things I look at, specifically, with software like this:

1. How widely supported out of the box is it? Is it mature?
2. How is the standard of code quality? Implementations etc?
3. What's the thing this software is best at?

We currently use a mix of PostgreSQL and MariaDB (a MySQL drop-in). Maria was chosen because MySQL is now an Oracle product and most of the cutting edge is with MariaDB in a mostly compatible project.

We have no need for horizontal scaling, and document storage like Mongo is simply not necessary. Mongo's license is also... problematic, from a lot of angles. Not just me. But NoSQL is fine, for specific things. It's not the right answer here.

I'll paraphase Mach_kernel, a former staff member here on the issues with MySQL, which I mostly agree with:

UTF8 encoding is actually not the full UTF8 plane, just non-4 byte characters. So UTF8MB4 was created as a workaround, and this was not a smart idea. All kinds of stuff broke because MySQL didn't bother doing its job right.

There's no multimaster replication in tree -- closest is Galera. This does matter for eventual plans.

InnoDB is a fragile box of compromises, and Aria isn't a suitable alternative. Postgres has very good robustness.

Anyways, next question.

2. Why not Xenforo? Why not use Apache? Why go for a custom stack?

Elf asked this many years ago, and because this was the early days of the site, well meaning users were inundating me, Praetor, and other volunteers with well-meaning suggestions. Here's the thing with me -- I don't do well under pressure, doubly when I'm having to process and understand thousands of opinions. It led to a lot of bad decision making in my early years of the site. But now I'm far less stressed, and a bit more experienced, so I'll put it to bed:

Xenforo is a nonstarter. It looks nice on paper, but proprietary board software has issues. Xen, VBulletin and other proprietary boards are encoded in Ioncube Loader -- in practical terms it means I'm at the mercy of the upstream to fix stuff. One thing I'm not letting happen in the future is to be in that bad position, I have devs for the web side who can help when my insufficient knowledge is not enough. Xenforo revokes licenses for petty reasons, too. It was never my thing. We started on PHPBB, the lack of support for agnostic protocols drove me to MyBB. MyBB ain't perfect, I know, but it is, well, excuse this but "Chabuduo, 差不多" Good enough. Here, I'm using it pragmatically.

Apache is... well a legacy piece of junk in my experience. Security and scaling issues, hard to configure,  breaks over little things... I was a cPanel admin for years and Apache is not great. I have considered Litespeed if we needed it but we don't. NGINX does okay.

As for why custom stack, e.g. FreeBSD and custom built packages, it's because I have to use GNU/Linux enough to know I don't want to do it in projects that don't need it. Does that make sense?

3. Why are so many things halfassed like the chat system?

Delicate balance. I don't want to use LDAP because complexity, so we're stuck with separate sign ons. I have an alternative idea using MyBB as a central authenticator though. More on that another time.

The chat stuff is a mess more generally because Mattermost took the manual account creation stuff away and I've not had time to look at alternatives.

4. What are the long term future plans for the site with the custom gallery coming, and beyond?

Everything will be FOSS, as usual, and designed to be compatible with a general Web 1.0 in the future, through many facets. Beyond the gallery, there will be more custom work to get projects more tightly integrated. The other things are eventual HA-type plans and a new marketplace, sorta like Dodoid's old IRIX One plan I suppose. The marketplace data (e.g. sale prices) will be eventually compiled into databases to show what kind of prices systems are going for, give ranges of prices etc. If other sites in the future want to contribute data to that project, I'll have guidelines.

Beyond that, more IRIX browsing compatibility, and I want to get the Nekotopia idea some more thought eventually.

Thanks guys, and feel free to ask questions or respond to anything I wrote. Just remember there are reasons behind what I did, even if you aren't happy with the reason.