terry@cssun.tamu.edu (Terry Escamilla) (07/14/90)
Someone please settle an argument we've been having at work. Is it possible for a workstation on an Ethernet to dynamically determine the Internet (or even Ethernet) addresses of the hosts/workstations immediately to its left and right? I guess this would be like an "intelligent" ping or ARP. Terry Escamilla terry@cs.tamu.edu
merlyn@iwarp.intel.com (Randal Schwartz) (07/14/90)
In article <23877@adm.BRL.MIL>, terry@cssun (Terry Escamilla) writes: | Is it possible for a workstation on an Ethernet to dynamically | determine the Internet (or even Ethernet) addresses of the | hosts/workstations immediately to its left and right? | I guess this would be like an "intelligent" ping or ARP. Ethernet is flat. (Well, OK, the cable is round. :-) Your broadcasts are flooded to every reciever on the cable. There's no left or right in electron space. Just another cybernetic hallucination, -- /=Randal L. Schwartz, Stonehenge Consulting Services (503)777-0095 ==========\ | on contract to Intel's iWarp project, Beaverton, Oregon, USA, Sol III | | merlyn@iwarp.intel.com ...!any-MX-mailer-like-uunet!iwarp.intel.com!merlyn | \=Cute Quote: "Welcome to Portland, Oregon, home of the California Raisins!"=/
stecz@hotwheel.dell.com (John Steczkowski) (07/16/90)
In article <23877@adm.BRL.MIL>, terry@cssun.tamu.edu (Terry Escamilla) writes: > > Someone please settle an argument we've been having at work. > > Is it possible for a workstation on an Ethernet to dynamically > determine the Internet (or even Ethernet) addresses of the > hosts/workstations immediately to its left and right? > I guess this would be like an "intelligent" ping or ARP. I believe you use /dev/right to find the ethernet id for the machine to your right, but to get the ethernet id for the machine on the left, you need the new GNU addon device driver /dev/ambidextrous. Implementation of these devices is hemisphere dependent, the default is for the northern hemisphere (except of course for the UK, since there internet addresses are backwards). John Steczkowski Dell Unix Customer Support P#: (512) 343-3571 9505 Arboretum Blvd. !s: uunet!dell!hotwheel!stecz Austin, Texas 78759 @s: stecz@hotwheel.dell.com
jrh@mustang.dell.com (James Howard) (07/16/90)
In article <7490@uudell.dell.com>, stecz@hotwheel.dell.com (John Steczkowski) writes: > In article <23877@adm.BRL.MIL>, terry@cssun.tamu.edu (Terry Escamilla) writes: > > > > Someone please settle an argument we've been having at work. > > > > Is it possible for a workstation on an Ethernet to dynamically > > determine the Internet (or even Ethernet) addresses of the > > hosts/workstations immediately to its left and right? > > I guess this would be like an "intelligent" ping or ARP. > > I believe you use /dev/right to find the ethernet id for the machine to > your right, but to get the ethernet id for the machine on the left, you > need the new GNU addon device driver /dev/ambidextrous. Implementation of > these devices is hemisphere dependent, the default is for the northern > hemisphere (except of course for the UK, since there internet addresses are > backwards). I thought there was a new 'proximity calculating protocol' (PCP) available that will attempt to determine the nodes immediately adjacent to the local machine. First of all, you must be running with a nameserver, which is used as a reference point. You must know where the nameserver machine is located relative to your own system. Then, go edit the space.c file under /etc/conf/pack.d/pcp and uncomment out one of the following lines as appropriate: /* #define DIR_SERVER LEFT */ /* #define DIR_SERVER RIGHT */ Then, rebuild a kernel. At which point, the driver executes 'ping' on each host recognized by the nameserver and saves the average round-trip time in ms. Then, the shortest two values are returned as the most nearly adjacent systems available. At this point, the definition of DIR_SERVER above becomes important. For example, if DIR_SERVER = LEFT, then the lowest time is assumed to belong to the machine closest to the nameserver, hence on the left. In the alternate case, it must be on the right side. Of course, this is difficult if the shortest times are identical, or if the load average on the machines are not all identical, of if .... No one is really taking this seriously I hope... ;-) --------------------------------------------------------- James Howard Dell Computer Corp. jrh@mustang.dell.com The opinions expressed are my own, and not necessarily those of my employer. ---------------------------------------------------------
frank@hpuxa.ircc.ohio-state.edu (Frank G. Fiamingo) (07/17/90)
In article <1990Jul14.063620.4375@iwarp.intel.com>, merlyn@iwarp.intel.com (Randal Schwartz) writes: > In-Reply-To: terry@cssun.tamu.edu (Terry Escamilla) > > In article <23877@adm.BRL.MIL>, terry@cssun (Terry Escamilla) writes: > | Is it possible for a workstation on an Ethernet to dynamically > | determine the Internet (or even Ethernet) addresses of the > | hosts/workstations immediately to its left and right? > | I guess this would be like an "intelligent" ping or ARP. > > Ethernet is flat. (Well, OK, the cable is round. :-) Your broadcasts > are flooded to every reciever on the cable. There's no left or right > in electron space. > > Just another cybernetic hallucination, > -- While it's true that the broadcasts are flooded to all receivers on the cable, it's not truly "flat". It takes a definite time for the electron to traverse any length of the cable. Conceivably one could measure the time it takes for a signal to bounce back to the original machine and convert that to distance. Frank Frank Fiamingo frank@hpuxa.ircc.ohio-state.edu (614)292-4843