I run 52 boxes across 8 providers. Every single one has different community semantics. Contabo uses 64512:100 for local pref. InterServer uses 64512:100 to blackhole. Same number, opposite result. I keep a spreadsheet. Its 4 MB now.
Communities could be simple. Theyre not. Providers love it because we cant compare $/GB properly when the knobs are invisible.
I. The Problem with Communities (as I mentioned above)
As I mentioned above, BGP communities suffer from several structural deficiencies that merit extended analysis. First, the lack of standardization creates cascading failures in automation pipelines. Second, the documentation gap forces operators into trial-and-error configuration cycles. Third, the semantic overload of transitive vs non-transitive communities introduces state complexity that most routing daemons handle poorly.
II. Counterarguments and Their Limitations
Some defenders of communities argue that:
- They enable fine-grained traffic engineering
- They preserve backward compatibility
- They allow provider-specific innovations
III. My Preferred Alternative (as I mentioned above)
As I mentioned above, I prefer extensive use of AS_PATH prepending combined with well-documented prefix-lists. This approach, while not perfect, eliminates the opacity that communities introduce. I have deployed this across three autonomous systems with measurable improvements in troubleshooting time.
IV. Conclusion
The gaslighting framing, while hyperbolic, captures something real about operator experience.
You can see them in the routing console without touching PowerShell. This is easier than some Linux tools where youre grepping through `vtysh` output. Im not saying communities are good. Just that the tooling matters.
Is it?
My provider strip all communities I send. I go go check check my config for one hour. The communities is it? Gone gone. I call them. They say "we dont pass communities." I say why you let me configure. They say silence.
So I am coming to agree with OP. Gaslighting. Is it?
4 MB of communities or the whole spreadsheet
Same number opposite result is wild honestly
