echan@cadev6.intel.com (Eldon Chan ~ ) (02/23/91)
On page 23 of RFC1122 (Requirements for Internet Hosts--Communication Layers) under the DISCUSSION paragraph said this: " The ARP specification [LINK:2} suggests but "DOES NOT" require a timeout mechanism to invalidate cache entries when hosts change their Ethernet addresses " Would someone tell me what are implemented and how it is done on SunOS, Ultrix and other UNIX machines ? The RFC mentioned four mechanisms-- timeout, unicast poll, link-layer advise and high-layer advice. So, which OS is using which mechanisms ? Thanks in advance for your input. Eldon Chan Design Technology Intel Corporation echan@scdt.intel.com
VOLZ@process.com (Bernard E. Volz) (02/24/91)
In <2669@inews.intel.com> echan@cadev6.intel.com writes: > On page 23 of RFC1122 (Requirements for Internet Hosts--Communication > Layers) under the DISCUSSION paragraph said this: > > " The ARP specification [LINK:2} suggests but "DOES NOT" require a > timeout mechanism to invalidate cache entries when hosts change their > Ethernet addresses " > > Would someone tell me what are implemented and how it is done on SunOS, > Ultrix and other UNIX machines ? > > The RFC mentioned four mechanisms-- timeout, unicast poll, link-layer > advise and high-layer advice. So, which OS is using which mechanisms ? > > Thanks in advance for your input. > > Eldon Chan > Design Technology > Intel Corporation > echan@scdt.intel.com For Process Software Corporation's TCPware for VMS we flush an ARP cache entry if we haven't RECEIVED a packet from the <Internet-Address>,<Ethernet- Physical-Address> pair within a reasonable time period (typically 10 minutes).
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (02/24/91)
In article <2669@inews.intel.com>, echan@cadev6.intel.com (Eldon Chan ~ ) writes: > Would someone tell me what [ARP cache timeout mechanisms] are implemented > and how it is done on SunOS, > Ultrix and other UNIX machines ? The most reliable way to answer such a question is to check the source, when possible. That way you don't rely on hearsay and rumor. I suspect 4.* BSD was influential in most versions of SunOs and Ultrix. (Take that statement as hearsay or rumor). The 4.3BSD-network-release code is in arptimer() in .../netinet/if_ether.c in your favorite public source repository, such as UUNET. An official tape of the 4.3BSD network source is cheap, and is required reading for any serious student of TCP/IP. If you want a rumor of how 4.3BSD does it, mine is that it uses simple timers which delete completed entries 20 minutes after creation. Vernon Schryver, vjs@sgi.com