Agreed. I had a consulting gig once, actually doing cyber security for Meta. They made us take an automated training, part of which was listening to videos of Mark Zuckerberg talking unironically about how important privacy is to the culture of Meta. The thing is, they had no good mechanism for making sure you actually watched the video. You could just mute Mark and then keep an eye on the run time, because at the end there would be a quiz. Most of the quiz questions were super stupid intuitive like “A friend asks you to use your Meta access to do X to their profile for them, what should you do?” And then multiple choice, with a bunch of obvious bad answers like “Like just do it, it’s fine.”

Companies attacking security researchers always goes so well for them.

I have a friend who once brought back an “iPhone” from China for laughs.

Truth.

I feel like Calvin can probably be placated with Minecraft.

The most dad joke thing about this is how no one under 35 will get it.

The right is ramping up attacks on Wikipedia for “partisan bias” right now. If you think Wikipedia has partisan bias, it’s a really good indicator that your beliefs are out of line with reality.

My absolute favorite thing about LLMs is how when they do that they explain the problem with “my code” to me.

Too real.

Meh. I’ll be impressed when it plays Aria Math.

(j/k, this is awesome, I love it)

Second Forgejo. Easiest deploy I’ve ever done.

I’ve been doing this since 2001 and I’m a big fan of openresty.

Sure, you can. There are even some very legitimate use cases.

But for your goals as you’ve defined them, you’re kind of just creating extra layers of complication for yourself. Sure, you could store all your data in MySQL AND Mongo and write to both with every write operation. If you do this, only read from ONE. You could write methods to toggle back and forth for which one you’re reading from but WHY?

A common usecase for this kind of thing is using Mongo as an object cache. You pull some common class of thing out of SQL that involves like

SELECT

g.group_id,

g.name,

g.blahblahblah,

COALESCE(json_agg(u.*) FILTER (WHERE u.user_id IS NOT NULL), '[]') AS moderators

FROM groups g

LEFT JOIN group_moderators gm ON g.group_id = gm.group_id

LEFT JOIN users u ON gm.user_id = u.user_id

GROUP BY g.group_id, g.name, g.blahblahblah;

Now you have this complex query returning your group object, with this JSON blob in it that stores all the moderator IDs. Now… what you can do is cache that in Mongo. When the API asked for a group by group_id, it can check to see if it exists in Mongo and if it does it can just grab that and return it to the user. If it doesn’t, it can ask SQL for it, then cache that in Mongo. If a new moderator gets added, that operation can refresh the Mongo cache.

Now WHY would you do this?

• Maybe Mongo has better performance.
• Maybe you’re trying to make less requests to your SQL server.
• Maybe Mongo runs locally on the same host as your endpoint, but the SQL server is remote and you want to use less bandwidth (this can cause some silly problems when you have multiple end points just sayin - but those problems have known simple solutions).
• You think its cool and you’re down to do the extra work.
• Any combination of these.

Now, IMO Redis is a superior tool for this, but I HAVE heard of people using Mongo.

But this isn’t really your usecase is it?

It could be a good learning exercise to set up your database reads and writes to use both MySQL and MongoDB. I’d actually recommend using either PostGres or MariaDB over MySQL… You’re going to encounter PostGres a LOT more often in the wild, so getting comfortable with it is a good use of your time. If you REALLY want MySQL without the taint of Oracle, Maria has got you covered.

Either way writing methods to read / write from both databases will certainly help you learn about both kinds of databases. Just make sure that when your site is actually live, you’re treating one database as the source of truth and the other as a mirror.

Given the scope of your project, I bet you’ll find both databases are perfectly adequate to meet your needs. But there are reasons many people have strong preferences for truely relational databases and it sounds like you’re kind of discovering that on your own.

But, you should probably just not complicate things for yourself at this point

You’re at a point where you can say

“I’m gonna just use Mongo, but I want to build things in such a way that it’s possible to change my mind later” just write your data handler class in such a way that you can save yourself pain down the road.

For example, you might have a data::getObject method() or a data::createUser() method. If you write this intelligently now, you can even preemptively add something like

if(database==="mongo"){ <-do stuff to make a new user in mongo-> }else{ <-Throw an unknown database exception-> }

Now, when you come back to switch databases in the future, all you have to do is

if(database==="mongo"){ <-do stuff to make a new user in mongo-> elseif(database==="postgres"){ <-do stuff to make a new user in postgres-> else{ <-Throw an unknown database exception-> }

But an even better and more secure way to do this is to make your data class ignorant of what kind of database you’re using (at the method level) and just define your schema. If you’re using FastAPI (I personally loath working in Python, eff FastAPI, I roll my own APIs in Rust, Lua and PHP) it’s kind of set up to get you started on this quickly. You probably should use it, given where you’re at. There’s reasons it’s the most popular tool out there for doing this kind of thing and will teach you a lot about how to structure an API.

https://www.fastapitutorial.com/blog/pydantic-schemas-in-fastapi/

https://hevodata.com/learn/fastapi-mongodb/

Yeah! Postgres is great!

• Mutters something under his breath about MariaDB.

My experience working with a vibe coder hired by one of our clients is actually that’s it great for MAXIMUM VP, as in a viable product made up of just under 40k lines of typescript, plus 90+ Node libraries for an app that amounts to a login page, a Square payment gateway and a user settings page (it’s seriously just a signup page for a coastguard and weather alerts service that the rest of our team built in Python and Rust). It crashes if it can’t talk to a database server that hosts no actual databases. It crashes if it doesn’t have the Square API secrets as envars, but the LLM also hard coded them into the API calls. It actually crashes if you try to run it any way other than “npm run dev” (so I srsly set up a service that runs it as npm run dev, as the ubuntu user).

Well yeah, they can’t keep training their models if all the available data is slop from their models.

A future is coming where it can just be your job to let an LLM observe your life 24/7… And that might be the only job available.

The way laws and bylaws describe the jobs of CEOs and CFOs, the most qualified people to do those jobs are sociopaths. Empathy is practically a disqualifying personality trait.

I mean, I didn’t bother reading the OpenAI ToS or privacy policy.

I just assumed they were recording everything I did and would probably sell it given the opportunity.

Please don’t promote Red Cap politics disguised as software.

systemd’s networkd has a built-in DHCP server; check option ‘DHCPServer’ and section ‘DHCPServer’ for that (same man page as above).

Is that true in Debian? If so, cool. I did not know that.