Re: [Tsvwg] checksum CRC reference

Lyndon,

I think the checksum/CRC discussion is suffocating from too much
information :-) IMO, what we desperately need now is some selection
criteria (something like a requirement statement/document).

Seemingly, nobody likes option/negotiation-based solutions. So we must
make a selection.

We know now that there are multiple checksum-based and CRC-based
solutions to choose from.

We know now that the implementation cost of each of the candidates
varies widely and depends on different factors (e.g., 32bit-machine vs.
64bit-machine, software vs. hardware, big cache vs. small cache, etc).

We know now that there are many different error models that can be used
to evaluate the effectiveness of the candidates.

We know now that none of the candidates is the clear winner in all the
error-scenarios (the only thing seems sure is that CRC-based ones
perform better than the sum-based ones).

We know now that different application areas are looking for different
level of transport data integrity checking.

Whichever direction we take, we must keep in mind that SCTP is a general
purpose transport and our final decision should be balanced and should
not be biased towards any specific group of applications or platforms.

regards,
-Qiaobing

Lyndon Ong wrote:
> 
> Folks,
> 
> For reference, David Black has passed a pointer to
> http://ietf.org/internet-drafts/draft-sheinwald-iscsi-crc-00.txt for a good
> discussion of CRC codes and their characteristics.
> 
> I have to say, Doug's results were quite confusing, esp. where the Adler-32
> seemed to be performing much better than modified Adler.  I did note that
> CRC seemed to perform best...
> 
> Cheers,
> 
> L. Ong
> 
> _______________________________________________
> tsvwg mailing list
> tsvwg@ietf.org
> http://www1.ietf.org/mailman/listinfo/tsvwg

_______________________________________________
tsvwg mailing list
tsvwg@ietf.org
http://www1.ietf.org/mailman/listinfo/tsvwg