On Mon, May 11, 2009 at 5:14 PM, Dave Miner <Dave.Miner at sun.com> wrote: > Mike Gerdts wrote: >> >> On Sun, May 10, 2009 at 12:47 PM, Sarah Jelinek <Sarah.Jelinek at sun.com> >> wrote: > > ... >>> >>> -You want to query network information >> >> In particular, I want to be able to use protocols that are already >> present on my network. ?That is, EIGRP, CDP, LLDP, etc. can be helpful >> protocols to understand network topology. >> > > I haven't looked deeply at those protocols' details, but I would be > concerned in some cases about determinism of results. ?Dynamic discovery of > network topology based on passive listening can be fairly unreliable in my > experience, and having flakiness of network infrastructure turn into > unanticipated deployment results seems like a bad path to head down.
Agreed. The same problem exists to a certain degree for DHCP and mdns, right? I actually worry more about mdns doing the wrong thing (I can accidentally install from the wrong server) than having automatic network discovery doing the wrong thing. If I don't have really careful post-installation verification, I may not realize that I installed from the wrong server until I'm running the wrong bits in production. If the network discovery goes wrong, the applications just don't work. If applications don't work, it is hard to miss this in the pre-production time of a server's life. Non-deterministic behavior that leads to a defect that isn't detected before a server goes live is a very bad thing in certain regulated environments. -- Mike Gerdts http://mgerdts.blogspot.com/
