Document: draft-ietf-rmt-bb-fec-raptorq-04
    RaptorQ Forward Error Correction Scheme for Object Delivery
Reviewer: Joel M. Halpern
Review Date: 27-January-2011
IETF LC End Date: 09-Feb-2011
IESG Telechat date: N/A

Summary: This document is ready for publicaiton as a Proposed Standard.

Note: The reviewer did not attempt to verify the actual FEC behaviors, but did
focus on whether he believed an implementor could follow the instructions to
implement the algorithm.  The comments below should be considered if the
document is revised.

Major issues: None

Minor issues:
    It would probably be helpful in section 3.2.3 if there was some comment as
to why the number of source blocks is a 12 bit field when the last paragraph
of 3.2.2 says that the number of source blocks is limited to 256.
    The order of presentation of material in section 5.3 is somewhat
confusing.  Rather than changing it however, I think it woudl suffice if a
forward reference, in the text early in 5.3.3.3 is added, saying that the
algorithm describes here assumes that the C symobls have been calculated
according to the constraints in section 5.3.3.4.  (The challenge appears to be
a matter of providing sufficient clarity without over-constraining
implementors.)
    In describing the "First Phase" algorithm in section 5.4.2.2, the chosing
and arranging step seems to ensure that the first row of V has a 1 in the
first column of V.  Confusingly, the text then refers to the chosen row having
an entry alpha in the first column of V.  It may be that I misread the steps
(in which case we should check if clarification is needed), it may be that
alpha is indeed always 1, in which case we should say so, or it may be that
the "ones" in the earlier selection criteria is merely supposed to be
"non-zeros".

Nits/editorial comments:
    In 4.4.1.2 when defining the symbols IL, IS, JL, and JS, the formula for
JL gets split across a line boundary.  It is very easy for a reader to miss a
piece, and get the completely wrong formula.  (The split causes the bind to
invert the operator priority, if it does not completely miss the extra
operation.)