SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    RE: iSNS: Last Call comments



    Hi Folks,
    
    Here is the resolution of the technical comments that I posted
    for iSNS.
    
    Regards,
    Josh
    
    > -----Original Message-----
    > From: Joshua Tseng [mailto:jtseng@NishanSystems.com]
    > Sent: Sunday, October 13, 2002 9:52 AM
    > To: ips@ece.cmu.edu
    > Cc: 'Elizabeth Rodriguez'
    > Subject: iSNS: Last Call comments
    > 
    > 
    > Hi Folks,
    > 
    > Here is a summary of technical last call comments against the 
    > iSNS spec -13:
    > 
    > 1)  There needs to be a way to scope the DevGetNext message.  
    > Recommend
    > that non-zero-length TLV's may be included as the first 
    > operating attributes
    > of the DevGetNext message.  They would be used to scope the DevGetNext
    > message.  Zero-length TLV's that follow would specify the attributes
    > of the next object that need to be returned in the response 
    > message.  If
    > there are no non-zero-length TLV attributes, then the 
    > DevGetNext message
    > is unscoped, as specified in previous iSNS spec revisions.
    
    The following text has been added to section 6.6.5.3 to allow scoping
    of DevGetNext messages:
    
    Non-zero-length TLV attributes in the Operating Attributes are used to scope
    the DevGetNext message.  Only the next object with attribute values that
    match the non-zero-length TLV attributes SHALL be returned in the DevGetNext
    response message.
    
    Zero-length TLV attributes MUST be listed after non-zero-length attributes
    in the operating attributes of the DevGetNext request message.  Zero-length
    TLV attributes specify the attributes of the next object that are to be
    returned in the DevGetNext response message.
    
    > 
    > Also for the DevGetNext message, it is ambiguous as to 
    > whether attributes
    > of other objects other than the next object, can be returned 
    > in the message.
    
    The following text has been added to section 6.6.5.3 to remove this
    ambiguity.  Only attributes of the object specified by the Message Key
    may be listed in the operating attributes.
    
    The Operating Attributes can be used specify the scope of the DevGetNext
    request, and specify the attributes of the next object that are to be
    returned in the DevGetNext response message.  All operating attributes MUST
    be attributes of the object type identified by the Message Key.  For
    example, if the Message Key is an Entity_ID attribute, then the operating
    attributes MUST NOT contain attributes of Portals.
    
    > 
    > 2)  In order to better integrate with SNMP, we need a method to query
    > for the next available index attribute.  Recommend a virtual 
    > attribute for
    > Entity, Portal, and iSCSI Node that always returns the value 
    > of the next
    > available "Index" attribute for that object.  This virtual attribute
    > would only be used for queries, and it would be an error to attempt to
    > register a value for this attribute.
    
    The following "virtual" attributes have been added with the corresponding
    tag values:
    
    Entity Next Index         8
    Portal Next Index        24
    iSCSI Node Next Index    38
    
    > 
    > 3)  Why is the registration period only 2 bytes, when the attribute
    > length is 4 bytes?  Recommend that the registration period 
    > use all 4 bytes,
    > not just 2 bytes. 
    
    The spec has been changed so that all four bytes are used for the
    Registration Period.
    
    > 
    > 5)  Spec needs to say that the iSNS server may reject iSNS messages
    > with incompatible message version numbers in the iSNS message header.
    
    The following sentence has been added to section 6.1.1.
    
    The iSNS server MAY reject messages for iSNSP version numbers that it does
    not support.
    
    > 
    > 6)  Section 7.l2.5 should clarify that "Protocol Version" refers to
    > block storage protocols.
    
    The first sentence of 7.2.5 now refers to "block storage protocols".
    
    > 
    > 7)  Clarify that if a query requests an attribute for which the iSNS
    > server has no value, then the server SHALL not return the requested
    > attribute in the query response.  Such query and response messages
    > SHALL NOT be considered to be in error.
    
    The following text has been added to section 6.6.5.2:
    
    If a DevAttrQry message requests an attribute for which the iSNS server has
    no value, then the server SHALL NOT return the requested attribute in the
    query response.  Such query and response messages SHALL NOT be considered to
    be in error.
    
    > 
    > 8)  The format of the DevAttrRegRsp message in section 6.7.5.1 seems
    > inconsistent.  The message key attributes seems to contain the same
    > thing as the operating attributes.  The message key should contain the
    > same attribute as the original registration message.
    > 
    
    The second paragraph of section 6.7.5.1 has been replaced with the
    following:
    
    The Message Key in the DevAttrRegRsp message SHALL return the Message Key in
    the original registration message.  If the iSNS server assigned the Entity
    Identifier for a network entity, then the Message Key Attribute field SHALL
    contain the assigned Entity Identifier.
    


Home

Last updated: Wed Oct 23 10:19:19 2002
11968 messages in chronological order