This message is from the T13 list server.

> ...

More free, fast, good education ... thank you again!

> Is this line of inquiry for understanding,
> or because a real problem has been encountered?

Yes both.

> I've never seen a problem like this 

Me, I've repeatedly seen problems exactly like this with AtapiPio.  I find these 
troubles familiar, because I've seen them likewise in Scsi & FireWire & ftp & ....

I've never seen a trouble like this with a hard drive and Ata - not even when the Ata 
was composed by a FireWire/Usb2 bridge.  In the real world, what provokes these 
troubles is removable media, exchangeable heads, variable block sizes, and the use of 
commands not standardised before both the host & the device shipped.

> > Re: [t13] UDma count well not - 3 of 4 - real examples
> >  "Pat LaVarre" <[EMAIL PROTECTED]> 12/02/01 07:49AM

My post here of real world AtapiPio examples was subject/time-stamped as shown.  With 
each real example I included a conjecture of why a non-removable, x200 byte block, Ata 
hard drive would not have a problem just where Atapi/ Scsi/ etc. did.

I'm trying to understand how well the AtapiPio solutions will or will not port to 
AtapiUDma, based on the understanding that AtapiUDma today offers bursts of 33, 66, 
100, 133 ... whereas AtapiPio runs out of gas at 17MB/s.

I'm sensitive to burst rate just now in part because this past summer I was surprised 
to see a form of connection where 66MB/s bursts were necessary to sustain just 17MB/s 
of traffic.  I had wrongly guessed 33MB/s would be adequate to sustain 17MB/s.  This 
makes me wonder how little the 17MB/s bursts of Pio can sustain.

I'd say noone can implement the  bInterfaceProtocol x50 of 
<http://www.usb.org/developers/data/devclass/usbmassover_10.pdf> without addressing 
these issues.

(((Lots of people can ship stuff claiming to implement bInterfaceProtocol x50 without 
understanding these issues: any defect that doesn't crash the o.s. more often than it 
crashes already may go unnoticed.)))

Pat LaVarre


Subscribe/Unsubscribe instructions can be found at www.t13.org.

Reply via email to