Just when you thought you got used to the weirdnesses in the networking implementations, you get a curveball like this one. Life is never dull if you test network devices.

Before releasing netlab release 2.0, I ran the full suite of integration tests for all devices for which I have the images. Interestingly, most VXLAN tests failed for Cumulus Linux 4.x even though we haven’t touched that code for ages.

Next step: trying to figure out what changed. The configuration changes were minimal. Even worse, the failure was non-deterministic. Somehow, we managed to transform a Cumulus Linux 4.x VM into a Heisenberg switch.

Jo attempted to follow the vendor Kool-Aid recommendations and use NETCONF/YANG to configure network devices. Here’s what he found (slightly edited):

IMHO, the whole NETCONF ecosystem primarily suffers from a tooling problem. Or I haven’t found the right tools yet.

ncclient is (as you mentioned somewhere else) an underdocumented mess. And that undocumented part is not even up to date. The commit hash at the bottom of the docs page is from 2020… I am amazed how so many people got it working well enough to depend on it in their applications.

Jan Schaumann published an interesting blog post describing the circuitous journey a browser might take to figure out that it can use QUIC with a web server.

Now, if only there were a record in a distributed database telling the browser what the web server supports. Oh, wait… Not surprisingly, browser vendors don’t trust that data and have implemented a happy eyeballs-like protocol to decide between HTTPS over TCP and QUIC.

A reader of my blog pointed out the following minutiae hidden at the very bottom of the Cumulus Linux 5.13 What’s New document:

NVIDIA no longer releases Cumulus VX as a standalone image. To simulate a Cumulus Linux switch, use NVIDIA AIR.

And what is NVIDIA AIR?

When deep-diving into the confusing terminology of switching, routing, and bridging, I mentioned you could perform packet forwarding at different layers of a networking stack. In this blog post, we’ll explore what happens when we combine packet forwarding on multiple layers within a single network, resulting in multi-layer switching, where edge devices perform Layer n forwarding (usually Layer 3), and core devices perform Layer n-1 forwarding (typically Layer 2).

Each layer can use any forwarding paradigm you choose. However, since we generally use IP at Layer 3, edge devices typically perform hop-by-hop destination-based forwarding, while core devices can use alternative methods.

I thought I’ve seen it all, but the networking vendors (and their lack of testing) never cease to amaze me. Today’s special: ArubaCX software VXLAN implementation.

We decided it’s a good idea to rewrite the VXLAN integration tests to use one target device and one FRR container to test inter-vendor VXLAN interoperability. After all, what could possibly go wrong with a simple encapsulation format that could be described on a single page?

Everything worked fine (as expected), except for the ArubaCX VM (running release Virtual.10.15.1005, build ID AOS-CX:Virtual.10.15.1005:9d92f5caa6b6:202502181604), which failed every single test.

Sean Goedecke published an interesting compilation of practical advice for engineers. Not surprisingly, they include things like “focus on fundamentals” and “spend your working time doing things that are valuable to the company and your career” (OMG, does that really have to be said?).

Bonus point: a link to an article by Patrick McKenzie (of the Bits About Money fame) explaining why you SHOULD NOT call yourself a programmer (there goes the everyone should be a programmer gospel 😜).

In the last few days, I decided to check out how much better ChatGPT has gotten in the last year or two. I tried to be positive and was rewarded with some surprisingly good results. I even figured out I can use it to summarize my blog posts using prompts like this one:

Using solely the information from blog.ipspace.net, what can you tell me about running ospf over unnumbered interfaces

And then I asked it about unnumbered interfaces and IS-IS, and it all went sideways:

Continuing the discussion started by my Breaking APIs or Data Models Is a Cardinal Sin and Screen Scraping in 2025 blog posts, Dr. Tony Przygienda left another thoughtful comment worth reposting as a publicly visible blog post:

Having read your newest rant around my rant ;-} I can attest that you hit the nail on the very head in basically all you say:

• XML output big? yeah.
• JSON squishy syntax? yeah.
• SSH prioritization? You didn’t live it until you had a customer where a runaway python script generated 800+ XML netconf sessions pumping data ;-)

Ethan Banks created an excellent introductory Netlab - Automate Your Network Labs With YAML! video showing how easy it is to start and configure a container- or VM-based lab, and explaining the basic netlab commands you need to get started.

Thanks a million!

P.S.: If you’ve done something similar and I haven’t noticed it, please send me the link.

Hendrik left an interesting comment on my Running IS-IS over Unnumbered Ethernet Interfaces blog post:

FRRouting (Linux) with pure IS-IS, the only way it currently (10.3) works is to copy the loopback IPv4 address to the interfaces that you need to do IPv4 routing on. The OpenFabric (IS-IS “extension” draft) does support true unnumbered interfaces and routes IPv6.

Let’s unpack this. There are (at least) four reasons a router needs an address associated with an interface¹:

A few weeks ago, I was criticising Nokia’s unnecessary changes to the SR Linux configuration data model, so it’s only fair that I also publish a counterexample:

• On April 12th, SR Linux failed one of the netlab integration tests. We keep adding functionality to these tests as we discover edge cases we didn’t test before, so sometimes a device that passed the test before might fail the modified version.
• I opened a netlab issue, believing it might be a configuration error on our part.
• It quickly became evident that we’re dealing with an SR Linux bug, as the failure to apply routing policies was random.

I thought that was the end of the story and closed the issue, but then something truly amazing happened:

netlab uses point-to-point links provided by the underlying virtualization software to implement links with two nodes and Linux bridges to implement links with more than two nodes connected to them. That’s usually OK if you don’t care about the bridge implementation details, but what if you’d like to use a bridge (or a layer-2 switch if you happen to be of marketing persuasion) you’re familiar with?

You could always implement a bridged segment with a set of links connecting edge nodes to a VLAN-capable device. For example, you could use the following topology to connect two Linux hosts through a bridge running Arista EOS:

I was considering an AI add-on that would have access to the netlab documentation and help you figure out how to use it for a few years, but never got around to implementing it (and surprisingly, with all the AI hype out there, there were no volunteers submitting pull requests). A few weeks ago, someone suggested adding an MCP server as an interface to ipSpace.net content, but the discussion quickly devolved into vague ideas.

However, as ChatGPT now has access to the live Internet, I decided to try out whether it can get the job done with a bit of prompting.

TL&DR: After a hiccup, it worked surprisingly well.

Andrew Yourtchenko and Dr. Tony Przygienda left wonderful comments to my Screen Scraping in 2025 blog post, but unfortunately they prefer commenting on a closed platform with ephemeral content; the only way to make their thoughts available to a wider audience is by reposting them. Andrew first:

I keep saying CLI is an API. However, it is much simpler and an easier way to adapt to the changes, if these three conditions are met: