SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    iSCSI: draft-10 more editorial


    • To: "Julian Satran" <Julian_Satran@il.ibm.com>
    • Subject: iSCSI: draft-10 more editorial
    • From: "Martin, Nick" <Nick.Martin@compaq.com>
    • Date: Fri, 8 Feb 2002 21:15:30 -0600
    • Cc: <ips@ece.cmu.edu>
    • content-class: urn:content-classes:message
    • Content-Transfer-Encoding: 8bit
    • Content-Type: text/plain;charset="iso-8859-1"
    • Sender: owner-ips@ece.cmu.edu
    • Thread-Index: AcGxGAdXKN51+971SzutXYvwKW5rZA==
    • Thread-Topic: iSCSI: draft-10 more editorial

    Julian,
    
    The following are some more comments about the iSCSI draft 10 document.
    These are primarily editorial comments.  
    
    Thanks,
    Nick
    
    As previously noted, the section numbers and page numbers are different
    between the text and pdf versions of the document.  All my comments are
    based on the pdf version.
    
    Page 26 "Initiator that identifying elements, such as..."  The word
    "that" does not fit.  One of the words "unique", "defined", "specific"
    or "assigned" may have been intended, or perhaps it can just be removed.
    
    On page 33, The capitalization of the quoted keyword "None" is confusing
    "(literal constants which may include "None")".  I have previously
    believed that keywords were case sensitive and that "none" (not "None")
    was the correct string.  I note however that nearly all other keywords
    have been capitalized ("Reject", "NotUnderstood", etc.).  For
    consistency with other keywords perhaps we should use "None".  (As
    someone else noted, "yes" and "no" are also not currently capitalized.)
    
    On page 35, I presume the text "(up to the mode page limit)" is obsolete
    since iSCSI mode pages have been removed.
    
    On page 36 "It is also an error ..., than the SCSI limit for first
    burst."  Should "SCSI" not be "iSCSI"?
    
    In the next sentence, which mentions "MAY request to send data blocks
    and a first burst of any size;..."  This looks like it might have been a
    reference to 0 meaning any value which is now removed.  
    
    On page 44, the sentence "During login, each end of the iSCSI session
    specifies the maximum iSCSI PDU size it will accept."  This sounds like
    the PDU size could be negotiated differently in each direction.  This is
    not how I think it works.
    
    On page 87, one paragraph starts with "LS:When".  I presume the LS:
    should be removed.
    
    This may be more of an issue than editorial:
    On page 89, section 7.7 Negotiation Failures, I believe another reason
    needs to be listed.
    -	The text request was answered with Irrelevant.  
    Alternatively perhaps Irrelevant was added as a way to avoid sending a
    reject.  If a negotiation ending in Irrelevant is to be considered a
    successful negotiation, this needs to be clearly stated.
    
    On page 115, in "than no more data PDUs are expected", "than" should be
    "then".
    
    On page 124, in the section Referenced Task Tag, ABORT TASK is referred
    to as TASK ABORT.
    
    On page 134, in the section R2TSN, the maximum number of R2Ts is given
    in hexadecimal 0xffffffff.  In most other cases (like DataSN on page
    131) the limits are expressed in decimal 2**32 or 2**32-1.  The form
    0xffffffff is generally used when specifying a reserved value.
    
    On page 141, for the list of characters valid in a text value, I would
    suggest also allowing the character '@' as would be used in e-mail
    addresses.
    
    On page 141, it is explained that binary items may be expressed in
    decimal, hex, or Base64.  Are the protocol defined binary values, such
    as MaxRecvPDULength required to be accepted in all these formats?
    
    Request for clarification/restriction of text key=value continuations:
    Are there any special rules for text key=value pairs spanning PDUs?  Is
    a pair of length less than MaxRecvPDULength permitted to span PDUs?  Is
    the sender required to fill MaxRecvPDULength before continuing to the
    next PDU?  Can a key=value pair which will require continuation begin
    somewhere in the middle of a PDU text buffer due to other key=value
    pairs already in the buffer?  If so, can a key=value pair be continued
    before the "=" ?  Is the text buffer to be continued required to fill a
    multiple of 4 bytes in the current text buffer (no padding required)?  
    I would personally prefer to see key=value continuation restricted to
    cases where the key=value string exceeds MaxRecvPDULength.  In other
    cases when the next key=value will not fit in the remaining space in the
    current PDU, one could put the entire key=value in the next text PDU.
    This would eliminate the possibility that one could be expected to
    correctly parse small strings such as ImmediateData=yes or
    MaxBurstSize=16384 with PDU breaks within them
    (MaxBurstSize=16<PDU-break>384).
    
    On page 149, in the section Login Parameters, the text appears
    contradictory.  "All keys in Chapter 12, except for the X- extension
    format, MUST be supported by iSCSI initiators and targets.  Keys in
    Chapter 12, only need to be supported when the function to which they
    refer is mandatory to implement."
    
    On page 173, there is no section header (and number) for the beginning
    of CHAP protocol.
    
    On page 179, an example for TargetAlias uses an apostrophe (') although
    this is not among the permitted characters (on page 141).
    
    On page 181, InitialR2T has no section number.
    
    On page 184, LogoutLoginMaxTime has no section number.
    
    
    


Home

Last updated: Mon Feb 11 01:18:02 2002
8720 messages in chronological order