This message is from the T13 list server.
On Sun, 29 Jun 2003, Nathan Obr wrote: > This message is from the T13 list server. > > > Hale, > The 'right thing' for the disk drive behavior that I was > referring to is the assumption that after a write to the disk, any read > from the same address any time after the write, will produce the same > data. The presence or absence of a cache should not affect disk > read/write behavior. I don't think we need a standard, a spec or a > white paper for people to understand this. If one assumes on a read verify the data come from the platter and not a spoof for benchmarks. There was a time in the past two year in Longmont @ The Raintree when a change of behavior was adopted for this reason alone. > Drive manufacturer are free to implements something that doesn't > behave this way, but they will have a hard time getting a logo. We > haven't had a hard drive fail for a long time and that is why I say they > have done the right thing for years. Okay remove the logo issue, and operate like Bob G and I did in the past. We did pushed OS issues togather in the spirit of "doing the right thing". > FUA is not new to drive manufacturers. The proposal to add FUA > to T13 was made a year ago. The only issue that a hard drive > manufacturer has had was around releasing which has started this whole > thread. That issue was very quickly resolved. Interestingly, very > little noise about FUA has come from the hard drive manufacturers, > however, if you think that FUA as it is in the spec, I suggest you do > what Western Digital did and write a proposal to change it. Point taken and noted. All business, I like it! Cheers, Andre Hedrick LAD Storage Consulting Group > Nathan > > -----Original Message----- > From: Hale Landis [mailto:[EMAIL PROTECTED] > Sent: Friday, June 27, 2003 8:58 PM > To: T13 List Server > Cc: Nathan Obr > Subject: RE: [t13] Re: hmmm.. no comments? > > On Fri, 27 Jun 2003 13:44:26 -0700, Nathan Obr wrote: > >I do assume that drive manufacturers will do the right thing, but that > >is only because they have for years. > > Are you sure? What is the definition of "right thing"? Is it some > test Microsoft uses for logo certification? Is it just some feeling > that everyone is doing the "right thing" so there is no problem? I > don't understand why you seem to have no concerns about this. > > >The only requirement that a host > >has on a device with a cache is that the cache doesn't change the > >behavior of the device. > > What behavior? Where is that defined? Is there some standard for how > a disk drive is supposed to work? Is there a standard for how a disk > drive cache should work? If such standards exist I've never seen T13 > or ATA/ATAPI-x reference them. Does Microsoft have a document that > describes in detail how a disk drive works and/or how a disk drive > cache works? If so, can you share it with us? > > >Any caching scheme that returns data for a > >sector other than the last data written to that sector has a design > >mistake, > > Huh? It would not be a "design mistake" if the manufacturer product > specification defined that their drive did not operate in that > manner. And I can think of design reasons to implement a drive in a > manner that you would call a mistake. Sure that may not be a "good > design" but there is no standard that would say it was > invalid/illegal. If you or Microsoft think such implementations are > not possible and/or would never happen then I think you and Microsoft > need to learn more about how disk drives are operate and how they are > designed. > > >however, I don't believe any caching algorithm is considered > >"valid" to ATA/ATAPI as I am not aware that ATA comprehends caching. > > Hmmm... ATA/ATAPI-x define SET FEATURES commands to turn read/write > caching on/off and then there are the FLUSH CACHE commands. Maybe > ATA/ATAPI doesn't define how read/write cache should operate but > ATA/ATAPI does allow the host some control of this drive activity. > > >I believe that device caching behavior is outside of the scope of the > >ATA protocol and off topic for this forum. > > Perhaps, but again that just reinforces the need for an answer to my > questions above, especially the question of where is the > specification or standard or Microsoft document that defines how a > disk drive and/or a disk drive read/write cache are implemented. > > In another email to the list today you said FUA should not cause > hardware changes in a drive (I think that is what you said). Before > you assume FUA is so simple I think you need to talk to some disk > drive designers. > > Hale > > > > *** Hale Landis *** www.ata-atapi.com *** > > >
