RARP usage
Kanoj Sarcar <[email protected]>
| Newsgroups | gmane.ietf.ipoib |
|---|---|
| Message-ID | <[email protected]> |
Hi, In section 3.0 of http://www.ietf.org/internet-drafts/draft-ietf-ipoib-ip-over-infiniband-01.txt RARP is indicated as one of the valid traffic types. This might lead people to falsely believe that RARP needs to be implemented on IPoIB. In fact, this might not be possible, since there is no mechanism in IBA to allocate the same QPN to the same interface each time the machine reboots, thus the mac address defined in the draft is variable across reboots, and thus RARP does not make sense. As pointed out in section 2.1.1 in http://www.ietf.org/internet-drafts/draft-ietf-ipoib-dhcp-over-infiniband-05.txt some implementations might choose to preserve the same QPN for the same interface (refer to "... if the implementation is designed to keep this association constant across boots"), and only then RARP might make sense. I propose some verbiage to the effect of "Implementation of RARP is optional for IPoIB and is only suggested on implementations designed to keep the QPN (which is a part of the mac address) to network interface association constant across boots. If the implementation does support RARP, it must use the 'ETHER TYPE' specified in this document." be added in section 3.0 of the previous draft just after Table 1. Thanks. Kanoj