Time to live

Mattermost and Jitsi are used by around five people daily, and so on.

I've set the time to live to 60 seconds, the lowest my DNS provider supports, on all DNS entries besides the email ones.

I opted for the shortest TTL so that if I have to migrate a server or something, I don't have to wait long until the DNS for it updates to its new IP address.

Is there any downside with my approach, or any benefit for me to set longer TTLs on the DNS entries? Or would it just be nice to my DNS provider if a longer TTL means I send them fewer DNS queries per day?

It would literally just be a kindness to your DNS provider who really doesn't need that kindness.

DNS queries are incredibly cheap.

The value of having a long TTL is, I mean, in theory, it could mean that repeat visitors to your sites and services don't have to do as many DNS lookups, so they have a little bit less latency in the initial click through when they get to your site of your service.

But in actual practice, that really isn't going to matter.

If I want to see the actual physical path my data takes to reach their perimeter, I'm gonna use traceroute.

Yes, and traceroute is fascinating because it exploits a built-in safety mechanism of the Internet called time to live, or TTL.

Break that mechanism down for us, because when I run a traceroute, it just lists out every single router my data bounces through, millisecond by millisecond How does my computer actually know the IP addresses of all those random intermediate routers sitting in different states or even countries?

Well, when you send a packet of data across the internet, there is a serious risk it could get stuck in an infinite loop just bouncing between two confused routers forever.

Right.

Exactly.

So to prevent the internet from clogging up with these looping ghosts, every single packet is assigned a TTL.

It's a maximum number of hops it is allowed to make.

It's actually in the metadata baked right into the packet itself.

Specifically the time to live or TTL value, right?

Exactly that.

When a packet travels across a complex network, it rarely goes straight from point A to point B.

Right.

Every single time that packet makes a hop through a routing device, its time to live counter is mathematically decremented by one.

By analyzing the final TTL of the returning packet and combining that with traceroute commands that force every router along the path to identify itself, well, these scanning engines can mathematically reconstruct the physical distance and logical layout of the entire network infrastructure.

That's brilliant.