tmt@osf.org (04/05/90)
> jon postel > the code in question is registered (91 = LARP), see RFC-1060 page 8. > However, i'd really like to see an explanation of why any machines crash > when a datagram with a protocol type they didn't know about before comes > along. Anyone have the story? I'd say the story is that the receiver sees an ethertype which looks like an 802.3 length field, goes off to look for the SSAP and DSAP or 91 bytes of packet and gets very confused. Granted the machine shouldn't crash! Most HP's support both Ethernet type 2 and 802.3 framing, and the fact that the patch is in ip_input is most curious. Apparently the kernel is handing up the LARP packet to the IP handler, not a good idea. Perhaps the IANA or someone can provide the story on the history of the LARP type? For instance why it's using a value in conflict with 802.3? Tom Talpey tmt@osf.org
postel@VENERA.ISI.EDU (04/05/90)
Lee: While LARP has an assigned number listed in RFC-1060, i'd not call it documented, perhaps identified. Since it is now out causing the public trouble it would be helpful if the authors of LARP did document it. --jon.
cire@CISCO.COM (cire|eric) (04/07/90)
Tom, This isn't an Ether type code but an IP Protocol type. So there isn't any confusion about a 91 byte length packet. Don't confuse protocol type with packet type. -c