SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    RE: iSCSI base64 and 12-92



    On Thursday, May 23, 2002 7:07 AM, Julian Satran wrote:
    
    > As we have to keep them for numbers I reinstated them for small values too
    > (why implement a check) and for those that don't take
    > numerical big-endian representation for granted there is now text that
    > says it  without referring to internal representation.
    
    Retaining base-64 encoding for normal-sized numbers (size less than or equal
    to the machine's largest arithmetic word size, usually 32 or 64 bits) may
    make the language in the spec simpler, but it does NOT simplify
    implementation by eliminating a check; in fact, it would require additional
    checks and implementation complexity.
    
    Keys that can contain large numbers (larger than the machine's largest
    arithmetic word size) must be handled differently from normal-sized numbers
    anyway, precisely because they are beyond the machine's normal arithmetic
    capabilities; therefore, a check based on the key type must be done no
    matter what.  If the spec requires support for base-64 encoding for
    normal-sized numbers, implementers will also have to check for the "0b"
    prefix, pass the string off to a base-64 decoding routine (taking care not
    to overflow the machine's word size), and then properly pad, byte-swap, and
    cast the result, or, if "0b" was not found, call strtol() instead.  This
    seems like needless implementation complexity to support something NOBODY
    will do, i.e., specify a normal-sized number like MaxConnections in base 64!
    If the base-64 requirement for normal-sized numbers were dropped, a single
    call to strtol() would suffice.
    
    In short, let's keep it simple: decimal and hexadecimal representations for
    normal-sized numbers, hexadecimal and base-64 representations for binary
    items and large-sized numbers.  (I would still argue that "large-sized
    numbers" are just a special case of binary items, but that's a fine point.)
    
    Michael
    
    --
    Michael J. Krueger              mailto:michael.krueger@windriver.com
    Wind River Networks                         http://www.windriver.com
    500 Wind River Way                               phone: 510-749-2130
    Alameda, CA  94501                                 fax: 510-749-2010
    
    


Home

Last updated: Thu May 23 16:18:34 2002
10264 messages in chronological order