@mitchellh I was a bit reluctant to move out from my previous terminal, and when tried Ghostty for the first time I didnt have so much feeling with it. Now it's my main driver and I'm very happy with it, so many thanks for the work @mitchellh
@Brady_H Same as you, never tested an Adidas before, mines only have 100kms and they are replacements for my 2nd NB Rebels v4. Good all rounders with the only point that the upper makes me small blisters in the right toes.
Following the Tour de France? Our guide covers all the stages with the likely decisive points identified, climb profiles, weather forecast overlay is coming soon! #TDF2026ᴿᴾ https://t.co/GLt8oS2Qnc
@DanielLockyer I was also a bit dissapointed with my values when they were measured, but after reading about it, VO2Max is not the only deterministic factor for performance. Running economy, lactate threshold, weight, etc are huge params to consider, and can be improved!
Me cansé de hacer la compra desde la web de Mercadona.
Así que convertí la API de Mercadona en una CLI.
Ahora cualquier agente de IA (o tú) puede buscar productos, gestionar el carrito y automatizar la compra desde la terminal.
Es open source:
https://t.co/XlW9qJn8UH
Just built automatic MTU sizing for Uncloud's WireGuard mesh so the cross-machine container communication never stalls.
WireGuard sets DF (Don't Fragment) bit on its own (outer) packets so setting the correct MTU on the link is important to avoid weird connectivity issues (smaller packets deliver, larger ones silently drop).
MTU = the biggest packet a link can carry, usually 1500 bytes. Every tunnel steals some of this for its own headers. WireGuard takes 60 bytes over IPv4, 80 over IPv6. So inside a WG tunnel on a 1500 link your real MTU is 1420 (worst case).
That's the correct tunnel size to ensure that everything that fits it gets delivered.
You can still send 1500-byte packets over a 1420 tunnel, and they get fragmented to fit or dropped. TCP can recover via PMTUD (path MTU discovery) but it relies on ICMP which is often blocked.
So you ideally want to right-size the packets routed through the tunnel at the origin. In our case that's the container packets coming from the Docker bridge network.
By default, docker network mtu is 1500. Note that setting a Docker bridge's MTU does nothing to your containers. eth0 inside a container is sized from the network's driver.mtu option. And it's immutable so you need to disconnect/remove all containers and recreate the network to change it.
Uncloud now detects each machine's egress MTU, subtracts WireGuard worst-case 80 bytes, caps at 1500, and floors at 1280 (IPv6 min). Then sets this as the MTU for both the WG tunnels and Docker networks so container traffic fits precisely into the tunnel. There's also a --wg-mtu flag if you want to override.
This is very similar to what Flannel and Cilium do by default: pod MTU = link MTU minus tunnel overhead.
It slightly caps internet egress too but the cost is ~0.15% throughput and ~5.8% more packets. Which is a fair price to make cross-machine traffic correct for both TCP and UDP.