Re: Review of requirements document

[Lani <lani@gadzoox.com>]

I agree that the spec appears very disk tape specific, and perhaps the confusing point here is that  we are doing a transport for all SCSI devices, but tend to describe storage concerns in particular detail. I do not think that the VBS point is worth lengthy debate, and am willing to accept the paper as is, but will offer the following solution by avoiding the use of all terms such as block,sector,library, etc. In the  first use of the term we are trying to distinguish file oriented storage from block or scsi oriented storage which may be a good focus point to keep, while the remaining two references in the spec could do with the general term of SCSI device... so here are the suggested sentence changes I would propose:

was: Traditionally, volume block...VBS....through fibre channel.
new: Traditionally, storage controllers utilizing the SCSI protocol have been attached to computers through the SCSI parallel bus or through Fibre Channel.

was: The IP/Ethernet infrastructure...VBS....current approaches.
new: The IP/Ethernet infrastructure offers compelling advantages for SCSI device attachment as compared to current approaches.

was: The iSCSI standard ....VBS...such as Ethernet.
new: The iSCSI standard will permit SCSI devices to be directly attached to IP networks such as the Ethernet.

-Lani



John Matze wrote:
> 
> The SCSI spec does use the sector representation in the Mode Select/Mode
> Sense pages for the format device page.  However, my request was more for
> SCSI software developers who in many cases have treated the blocks as
> sectors.  In some cases it maps easier with the upper layers of the
> operating system.  In either case if this sounds too confusing, then I have
> no objection of dropping the term.
> 
> This does bring up another objection if we do change the VBS.  Not all SCSI
> devices are block and volume oriented.  For instance tape libraries do not
> have a concept of blocks they use elements and I'm not sure about scanners
> and printers.  I wouldn't want the spec to be too "disk/tape" specific.  It
> may stop other devices from utilizing the spec.
> 
> John
> 
> -----Original Message-----
> From: Lani [mailto:lani@gadzoox.com]
> Sent: Thursday, June 29, 2000 1:27 PM
> To: John Matze
> Cc: scsi-tcp@external.cisco.com; ips@ece.cmu.edu
> Subject: Review of requirements document
> 
> John,
> In reviewing Randy's requirements document in our phone meeting today
> (6/29/00) a change was proposed/introduced (starting on first page under
> section APPLICABILITY) to define current storage controllers ( I assume
> SCSI) as volume/block/sector(VBS) oriented devices & I felt that this was
> confusing to add "sector" to this definition, as the SCSI protocols used on
> such storage devices are block (or stream) oriented and the specs never use
> the term "sector". We needed input from you as this change was initiated
> from your comment at the 6/19 Haifa meeting. Could you please elaborate on
> what you had in mind - should we attempt to describe the media mapping
> method underneath SCSI in the document?
> thanks - Lani