RE: DDP segment sequences for Untagged DDP message

At 08:01 AM 1/27/2004 -0600, Culley, Paul wrote:
>Note: this is the second sending, it did not show after the first time
>within 24 hours...
>
>OK, I realize that I missed a small point; in the first case below
>
>   1. QN = 0, MSN = 1, MO = 100 L = 1 Payload = 100 bytes
>   2. QN = 0, MSN = 1, MO = 0   L = 0 Payload = 100 bytes
>   Result: DDP Message (bytes 0-200) is delivered to ULP
>   [prc] yes
>   [pat]DDP message (bytes 0-200) occurs after 1.
>      2 will cause an invalid MSN error
>
>I was thinking of TCP segments as they arrive at the receiver, not as
>sent at the transmitter.  If the example was sent in this order, I agree
>that an invalid MSN error is the correct answer, if they were merely
>received in this order (they got flipped in route and were received out
>of order, but sent in order), then the result would be a normal
>delivery.
>
>I am not certain I agree with Pat on the "optional" nature of checking
>for errors that are not even mentioned in the specification.  I think
>that if we have to list all of the checks that we DO NOT perform in the
>specifications, then we have quite a bit of writing left to do.

It would be both impractical to document all that should not be done and a 
pain / low ROI to test for compliance.


>That said, we should perhaps be a bit more clear about the rules on
>Sending overlapped segments, or Sending Gaps.  I am definitely not in
>favor of requiring the receiver to check for either case.

Concur.

>If we were to add text to DDP for this I think it would say something like:
>
>When sending an Untagged Message, the Sender MUST send data to each
>offset within the target buffer only one time.
>
>When sending an Untagged Message, the Sender MUST send a contiguous
>stream of bytes to the destination buffer, starting at offset 0.
>
>In my mind, we should allow the receiver to check for these conditions
>only if these rules are made on the sender.

I don't think we need to add these.

Mike



>Paul R. Culley
>HP Fellow
>281-514-5543
>
>
>-----Original Message-----
>From: [email protected] [mailto:[email protected]]
>Sent: Friday, January 23, 2004 4:45 PM
>To: Culley, Paul; [email protected]; [email protected]
>Subject: RE: [rddp] DDP segment sequences for Untagged DDP message
>
>
>My responses are marked pat
>
>-----Original Message-----
>From: Culley, Paul [mailto:[email protected]]
>Sent: Friday, January 16, 2004 12:28 PM
>To: Barry Reinhold; RDDP
>Subject: RE: [rddp] DDP segment sequences for Untagged DDP message
>
>
>Below...
>
>Paul R. Culley
>HP Fellow
>281-514-5543
>
>
>-----Original Message-----
>From: [email protected] [mailto:[email protected]] On Behalf Of
>Barry Reinhold
>Sent: Friday, January 16, 2004 1:09 PM
>To: RDDP
>Subject: [rddp] DDP segment sequences for Untagged DDP message
>
>
>
>So, is there agreement, for the sequences that follow, that the
>specified "result" given below is the correct (and desired?) behavior as
>defined by the 01 draft.
>
>Untagged DDP Segment Sequences:
>
>1. QN = 0, MSN = 1, MO = 100 L = 1 Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 0   L = 0 Payload = 100 bytes
>Result: DDP Message (bytes 0-200) is delivered to ULP
>[prc] yes
>[pat]DDP message (bytes 0-200) occurs after 1.
>      2 will cause an invalid MSN error
>
>1. QN = 0, MSN = 1, MO = 100, L = 0, Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 0,   L = 1, Payload = 100 bytes
>Result: ULP sent Error type=Untagged Buffer Error, Error Code=Invalid MO
>[prc] No, the receiver will place both segments, but inform the ULP that
>only 100 bytes was delivered, ending at MO=99.  No error should be
>detected or reported, unless the receiver's buffer was not large enough
>to hold the second 100 bytes. [pat] Checking for the error of last
>segment not containing highest MO is optional. Both behaviors above
>would be valid responses.
>
>1. QN = 0, MSN = 1, MO = 0, L = 0, Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 0, L = 1, Payload = 100 bytes
>Result: DDP Message (bytes 0-100) is delivered to ULP
>[prc] yes, but it is not certain which segment ends up in the buffer, it
>won't matter unless the two payloads are different. [pat] agree.
>
>1. QN = 0, MSN = 1, MO = 000, L = 0, Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 200, L = 1, Payload = 100 bytes
>Result: DDP Message (bytes 0-300) is delivered to ULP
>[prc] yes, but the middle 100 will contain unknown values (probably what
>they had before the buffer went through DDP). [pat] There are two valid
>behaviors. The receiver may deliver the message as above or it may be
>checking and report an error.
>
>1. QN = 0, MSN = 1, MO = 300, L = 1, Payload = 100 bytes
>Result: DDP Message (bytes 0-400) is delivered to ULP
>(This is based on wording in 6.3, MO field. Note that the second
>sentence simply says that the MO "that references the first octet MUST
>be set to zero", not that a segment having a zero MO is required) [prc]
>yes, but first 300 is unknown [pat] Again the receiver is allowed to
>deliver or to report an error.
>
>1. QN = 0, MSN = 1, MO = 100, L = 0, Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 000, L = 1, Payload = 0 bytes
>Result: ULP sent Error type=Untagged Buffer Error, Error Code = Invalid
>MO (There is ambiguity between 6.1 L bit field description ( Segment w/
>L bit set must have highest MO), and 9.1 (implies that for non-zero
>length segments the header fields are not validated.) [prc] no, at least
>not because of a protocol error.  There is only an error if the buffer
>limits as defined by the ULP is exceeded.  You should get an indication
>that a zero length message was delivered. [pat] The ambiguity you point
>out is probably an error. The MUST on checking Untagged DDP segments
>required that they be validated before placement. A zero length segment
>isn't placed so the MUST didn't apply to it. There probably should be a
>statement about checking the fields before delivery of zero length
>untagged messages or zero length segments with the L bit set. Also even
>if that statement is added, this is another case where the receiver is
>allowed to deliver or report an error.
>
>1. QN = 0, MSN = 1, MO = 000, L = 0, Payload = 100 bytes
>2. QN = 0, MSN = 1, MO = 100, L = 1, Payload = 0 bytes
>Result: DDP Message (bytes 0 - 100) is delivered to ULP
>[prc] yes
>[pat] yes. Note that this is another case where there should be a
>requirement that the zero length segment header values be checked before
>delivery.
>
>1. QN = 0, MSN = 1, MO = 100, L = 1, Payload = 0 bytes
>Result: DDP Message (bytes 0 - 100) id delivered to ULP.
>[prc], yes, but with unknown data values.
>[pat] May deliver the message or may scoreboard and report an error.
>
>Barry Reinhold
>Lamprey Networks
>[email protected]
>(603) 868-8411
>
>
>
>
>_______________________________________________
>rddp mailing list
>[email protected]
>https://www1.ietf.org/mailman/listinfo/rddp
>
>_______________________________________________
>rddp mailing list
>[email protected]
>https://www1.ietf.org/mailman/listinfo/rddp
>
>_______________________________________________
>rddp mailing list
>[email protected]
>https://www1.ietf.org/mailman/listinfo/rddp