On Mon, Nov 04, 2024 at 03:34:55PM +0200, Saku Ytti wrote:
> I wish this would have good outcomes, but almost no customers use
> advanced features, which cost money to develop and maintain.
>
> Likely by voting with your feet, support expensive customers aggregate
> to feature full companies, and support cheap customers aggregate to
> feature empty companies, creating perverse incentive, especially in
> such markets as IP transit, which is largely seen as interchangeable
> and the only metric is cost.

        Yes, the challenge I've seen here is that as randy said "so give
me free transit", the issue that i've seen is propogation of routes much
further than expected which makes things far more difficult, and this
gets worse when you talk about any sort of anycast(ed) prefixes and how
you monitor those.

        It takes a lot of time as well to debug and figure out if it's
an oddity of the routing architecture or something simpler that leads to
these issues.  I also know that testing all these variants can be quite
hard as well, so there's something to be said about the investment
point.  Also saku is right on the below as well:

> Nanog is unfortunately not representative.

        - Jared

--
Jared Mauch  | pgp key available via finger from jared@puck.nether.net
clue++;      | http://puck.nether.net/~jared/  My statements are only mine.