RE: iSCSI: Reliability of markers (was Re: R2TDataSN and other recovery mechanisms)

["KRUEGER,MARJORIE (HP-Roseville,ex1)" <marjorie_krueger@hp.com>]

> Not sure what you mean "current marker approach"?  The 
> appendix in iSCSI 05.txt lists two approaches, clearly the 
> second marker approach (fixed intervals) doesn't suffer from 
> this disadvantage.  In any case, a marker scheme would be 
> roughly between the iSCSI layer and the TCP layer, so it's 
> not appropriate to include in the iSCSI header digest.
> 

Oops, I again re-read appendix C, it appears to have changed from my
original understanding of "markers at fixed intervals" and as specified it
would suffer from the same undetected corruption if not protected.  However,
this approach would not suffer greatly from unprotected marker information,
since an invalid marker could just cause an implementation to skip ahead to
the next marker and statistically, one should eventually be right ;-)

Cheers,
Marj