Re: LSF Papers online?

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2009-04-15 at 00:09 +0200, Bartlomiej Zolnierkiewicz wrote:

OK, look, guys, could we stop this argument?  It's becoming a bit old.

For the record, I was unhappy to have libata pata in SCSI, but it got
moved into drivers/ata shortly after, limiting my influence.  I thought
having a slow wind down of legacy pata in drivers/ide but moving to
libata for sata was the correct split.  I agree with Jeff that the SCSI
layer did provide unique features that SATA/NCQ needed at the time ...
but I also think we need to move them into block sooner rather than
later.

I'm the last person ever to proscribe a kernel subsystem that has
willing volunteers, so drivers/ide is yours as long as you want to
maintain it.  I can certainly see the merits of a thinner stack to the
embedded world, and not a few embedded developers seem to agree.

As far as moving it out of SCSI goes, I've always been a supporter of
this.  Tejun looks like he's willing to execute, so I feel much more
sanguine that it will happen.  The point of the notes was really to draw
attention to something I hadn't realised:  we can't refactor block to
get libata out of SCSI without also changing drivers/ide, which does
make the problem harder.

On a final note about the urgency of getting libata out of SCSI: Intel
has been worrying for a while about the fatness of the SCSI/libata
stack, and its effects on performance, especially command transmission
via SAT, so I'm hoping they'll be supporting the effort.

Jmaes