Miles Nordin writes:
> --pgp-sign-Multipart_Sun_Jan__8_14:16:11_2006-1
> Content-Type: text/plain; charset=US-ASCII
> 
> >>>>> "bh" == Brian Hechinger <wonko@4amlunch.net> writes:
> 
>     bh> opinions?
> 
> opinion is that any hardware RAID that does not include battery-backed
> NVRAM is a fraud and may actually be worse than an unRAIDed disk (and
> this assertion does kind of agree with what you hear other people
> mumbling about weird IDE raid disasters)
> 
> this is kind of zealous Sun propaganda, but it's the only explanation
> of the ``RAID5 Write Hole'' I've stumbled upon so far:
> 
>  http://blogs.sun.com/roller/page/bonwick?entry=raid_z
> 
> I'm not sure how RAIDframe addresses this issue, if it does a better
> or worse job than the cheap chinese cards?

Well... RAIDframe tries to keep track of whether or not it *knows* 
the parity is in-sync with the data.  If the condition of all parity 
bits are not known, that's where the "raidctl -P" after a reboot 
comes in... (it will verify that all the parity bits are in sync with 
the data, and correct any parity bits that are not. )

Later...

Greg Oster