[jeffbee]

I would say the bullet points at the top are not strictly correct. The response does not necessarily transit the proxy. Responses can be returned directly to the client (DSR).

[nyrikki]

> Note: For simplicity, we’ll focus on Layer 7 (HTTP) reverse proxy.

Layer 4 proxies are a very specific sometimes food that most people should actively avoid until they need it because of the tradeoffs.

DSR is layer 4, and not in scope of this post.

[jeffbee]

Your comment, to me, only points out that the OSI layer model is nonsense. Envoy in DSR mode routes traffic based on application features, at "layer 7".

[nyrikki]

Envoy calls it Layer 4

https://blog.envoyproxy.io/introduction-to-modern-network-lo...

[jeffbee]

That's fair. Of course that post also calls the OSI model "unfortunate" and "a poor approximation".

[nyrikki]

All models are wrong, some are useful.

Layer 4 to 7 is useful in this case, as layer 4 involves forging tcp/udp packets, which is vastly different than say a http level reverse proxy.

[thequux]

If you actually read ITU T-REC X.200, which specifies the OSI model, you'll find that it doesn't match the modern internet at all. E.g., we don't have an OSI-style transport protocol at all (connections themselves aren't addressable independent of the SSAPs), TCP and UDP are actually layer 5, the presentation layer is protocol-specific, and pretty much the entire stack falls to bits if the network layer isn't packet switched.

There's a separate term for the bits of the OSI model that are actually relevant; it's called the IETF model.

[philwelch]

The model itself isn’t nonsense because it’s not a model of load balancers; it’s a model of network protocols. Load balancers might handle multiple levels of the stack for the same traffic, but so does any other networked program, eg handling cross-domain redirects.