Jeffrey Haas wrote on 16/04/2026 14:20:
Namely, that prepending is helpful for traffic engineering but that if you do too much of it, there are gotchas.
yes, but from a practical point of view, "too much" is arguably "any at all".
I.e. there is no amount of prepending on the internet dfz which is safe, in the sense of being small enough that it wouldn't be an aggravating factor for the problems described in the document.
This opens up a secondary issue of how many upstream ASNs you want in your as-path. The answer is very few, if you want to view a short as-path as the primary mechanism for ensuring that the impact of prefix hijacking attempts is minimized, because from a hijacking point of view, one asn in a path is the same as another: it adds 1 to the overall metric in the distance vector calculation.
The consequence is that using as-path prepending for TE should arguably be deprecated unless there are other risk mitigation mechanisms in place, which in future would include ASPA.
Possibly this should be documented, but I don't think this there's enough material here to justify an RFC.
Nick _______________________________________________ GROW mailing list -- [email protected] To unsubscribe send an email to [email protected]
