When I write code for a feature in IOS-XR, I invent a CLI. I’d like to know what you guys who automate using the CLI think. What should I keep in mind? Here’s a few things I already consider: For a show command that has tabular output, sometimes we spill over into two lines for an object. Especially once IPv6 addresses were invented. I invented the “wide” keyword to show commands to keep all info for one object to a single line. But the line length can extend to infinity. If column widths need to change, because something gets wider, then I will change the width. Example, AS number going from 2 bytes to 4 bytes. For non-tabular output, everything I display should have some keyword that the automation looks for. We try to never change those keywords. We may add new keywords for new information or move them, but we try to minimize those movements. We try REALLY HARD for idempotency. We fail rarely. Kind Regards, Jakob ---------- Original message ---------- Date: Tue, 18 Mar 2025 17:56:33 -0400 From: Tom Beecher <beecher@beecher.cc> I grew up on the CLI, as many did, so I am most comfortable there. There have always been downsides when it comes to automation, as screen scraping sucks.