I believe that Melinda discusses a method for transmitting data thru
hostile gateways in her paper on Rita. 

-----Original Message-----
From: CMSTSO Pipelines Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of Paul Gilmartin
Sent: Tuesday, July 31, 2007 11:20 AM
To: [email protected]
Subject: [CMS-PIPELINES] Checksums? CRC? (and hex display?)

Looking for data validation, I find the CRC stage.  But 16-bit checksums
are much in disrepute -- probably good enough for detecting noise, but
not sound cryptographically.  (Heck, even the 128-bit MD5 is now
suspect.)
Is better available somewhere?  SHA-1?  Or even the lowly MD5?  Weakness
notwithstanding, MD5 has the advantage of being readily available on
other platforms.

And I readily suspected and proved a weakness in PIPE CRC: files
with content:

    a
    bc

and:

    ab
    c

show identical CRC checksums -- apparently the algorithm ignores record
boundaries.  This is unfortunate because such framing errors are readily
introduced by transfer to a stream-oriented workstation and back.  And
consider Melinda's(?) assertion that CMS Pipelines is superior to other
pipe utilities because not all data should be treated as streams.  I
suppose PIPE PACK | CRC can be used to overcome the record boundary
deficiency.

For my testing, I would have liked a stage comparable to the Rexx C2X
function.  Of course, it's easy to write, but has someone else already
invented that wheel?

Thanks,
gil
--
StorageTek
INFORMATION made POWERFUL

Reply via email to