|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] RE: iSCSI: Use of Reject as a key valueKevin, The current text says explicitly that reject is a valid response (i.e., not a protocol error). The addition makes only explicit the obvious. On the other hand a reject may be bad for some parameters (e.g., some locally enforced authentication). In that case you are free to close the session with a login error. But that IS NOT the general case. The wording about Reject not being a protocol error is NOT a change. Julo
The iSCSI spec did not specify the handling of reject prior to the current v15 working version. I am NOT proposing changing the case sensitivity. I am merely pointing out a potential problem in the proposed handling of reject during login negotiation. Our current implementation handles it as described by Professor Russell as "Case 2", the rejection of a parameter during login (except for markers) terminates the login. My example was meant to demonstrate the problem associated with continuing the login with rejected parameters. The initiator and target have the potential to continue with miss-matched parameters. Professor Russell had three proposals in what he posted. He was not requesting a particular wording but meant to open it up to the list for discussion and a solution. Kevin -----Original Message----- From: Julian Satran [mailto:Julian_Satran@il.ibm.com] Sent: Friday, July 26, 2002 1:33 AM To: kevin_lemay@agilent.com Cc: ips@ece.cmu.edu; Julian Satran; pat_thaler@agilent.com; rdr@io.iol.unh.edu Subject: RE: iSCSI: Use of Reject as a key value Do you propose the formulate the standard so that it can tolerate all initiator syntax errors? Or just we state that it should be case insensitive. The first would be a bad idea. The protocol has enough complexity in order to deal with transport errors. It would be a bad idea to add implementation errors to the list. If you want case insensitivity - this was discussed and rejected a long time ago - again because it felt unnecessary. I would also like to point out that this is not a change. This behavior was always assumed (and implied in the statements about negotiations atomicity). The statement was added at the request of Bob Russell as a clarification. Julo kevin_lemay@agile nt.com To: Julian Satran/Haifa/IBM@IBMIL, rdr@io.iol.unh.edu cc: ips@ece.cmu.edu, pat_thaler@agilent.com 07/25/2002 09:59 Subject: RE: iSCSI: Use of Reject as a key value PM Julian, I think that this change is a bad idea. The spec says that: "If the acceptor sends "Reject" as an answer the negotiated key is left at its current value (or default if no value was set). If the current value is not acceptable to the proposer on the connection or session it is sent the proposer MAY choose to terminate the connec- tion or session." But what happens in the following case: i->ImmediateData=Yes t->ImmediateData=no (Note: wrong case used) The initiator must use the default, which is ImmediateData=Yes, and there is no way for the initiator to inform the target of the error because sending: i->ImmediateData=Reject would be a re-negotiation. This will not work! Kevin -----Original Message----- From: Julian Satran [mailto:Julian_Satran@il.ibm.com] Sent: Wednesday, July 24, 2002 7:52 PM To: Robert D. Russell Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu Subject: Re: iSCSI: Use of Reject as a key value Bob, On the last pdf you will find a statement about what to do. Julo "Robert D. Russell" <rdr@io.iol.unh.edu> Sent by: owner-ips@ece.cmu.edu 07/25/2002 12:58 AM To: ips@ece.cmu.edu cc: Subject: iSCSI: Use of Reject as a key value Julian: Attached are 2 ascii text files. The first, reject_extracts.txt, contains all the pieces of draft 14 (the latest available txt version of the drafts) that say anything about the use of Reject as a key value. (At least all those I could find using simple search -- unfortunately, Reject is also the name of a PDU and hence it is not a simple mechanical process to distinguish the two uses! If I missed some, please let me know.) The second, reject_comments.txt, are my comments on these excerpts from the standard. It seems to me that the key thing missing in the standard is a general statement about "what to do next" if a key=Reject response is received. Except for the OFMarkInt and IFMarkInt keys, I could find no other statement about how to proceed after receiving the key=Reject response. In looking through the e-mails posted to the list for June and July, I also could find nothing, although many people seem to be taking the third of the 3 interpretations I listed in reject_comments.txt. I am requesting a clear statement somewhere in the standard that says "what to do next" upon receiving a key=Reject response. Thank you for your consideration. Bob Russell InterOperability Lab University of New Hampshire rdr@iol.unh.edu 603-862-3774 #### reject_extracts.txt has been removed from this note on July 25 2002 by Julian Satran #### reject_comments.txt has been removed from this note on July 25 2002 by Julian Satran
Home Last updated: Tue Jul 30 10:39:07 2002 11481 messages in chronological order |