On Wed, Mar 9, 2011 at 6:22 PM, Randy Bush <[email protected]> wrote: >>> I'm personally a fan of keeping things simple .... what's the beef >>> with tcp/22 here? >> The set of source IPs needing access to tcp/22 for mgmt may not be the >> same as the set of IPs needing access to tcp/22 for the rpki service > > the router is the originator of the ssh connection, so there is no guess > as to the port the cache will use. if you are a big fan of ip address > based acls, it is the cache you will be protecting, not the router.
apologies I reversed the info/pkt flow. Ok, so ideally in a sane world I can already scp/ssh from a router to another router (maintenance requirements + code moves + config loads ... last-resort 'ohshit' configs.) Adding a few cache ips where the only thing running on the ip/port combo is this service seems ok to me. Ideally cache box has 'service' ip and 'maintenance' ip... with different security policies and 'better' service separation than I can get today in the world of routers. I'm back to 'why complicate things?' -Chris > i think the issues were > o predictability and default config of router > o other tools may want to reach server > > on the ssh vs other transport issue, i am less happy. the protocol > specifies ssh, and this discussion would lead to standardizing the port > on the cache/server to be used. that the testbed currently provides a > few clear-text servers for implementors to test without ssh is merely a > test service, and, imiho, should not be standardized. unless operators > would care to argue that they trust security services which use > clear-text. > > randy > _______________________________________________ > sidr mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/sidr > _______________________________________________ sidr mailing list [email protected] https://www.ietf.org/mailman/listinfo/sidr
