LASER MAGNETIC STORAGE                            MEMO
     INTERNATIONAL COMPANY
_______________________________________________________________________

    Optical Storage Division
                                                     25 April 1988

To:       SCSI Standards Committee X3T9.2

From:     Paul Boulay
          (719) 593-4323     

Subject:  WORM Device Specific Mode Select Page

Write-Once-Read-Multiple-Devices today support three distinct 
markets: data archive, image storage and on-line reference. In some 
data archive uses, the integrity of the data at least partially 
depends on the writable only once characteristic of the media.

When the Update Block command is implemented on a WORM device used 
in this sort of application the extra security provided by the 
writable only once feature is compromised. There are several avenues 
available to a WORM device supplier to remedy this situation. The 
most obvious is simply to never implement the Update Block command. 
However, there are a few reasonable applications for the Update 
Block command. (These involve managing the media rather than 
modifying user data files.) 

The minimum level of support for an archive application which relies 
on the writable only once characteristic is to be informed if the 
data is suspect. A normal Read command should raise some sort of 
exception in this case. Many users will not need or want such an 
exception raised. Also, the same user who needs detection of updated 
blocks when reading one file may not want an exception raised when 
reading other files. 

If a device supplies both the Update Block and the Read Updated 
Block commands, a user has the information needed to recover from an 
unwanted update of a block. 

The SCSI implementation of these ideas involves creating a Mode 
Select Page specific to devices with WORM media. This would be 
devices using the Write-Once device type plus Optical Memory Devices 
which have WORM media installed.

This Mode Select Page would include a bit which would enable 
reporting instances of reading a block which had been updated.

If the Report Updated Block Read (RUBR) bit was set and a Read 
command encountered such a block, the Read would complete with a 
Check Condition. A Recovered Error Sense Key would be posted and 
additional sense codes would indicate this case. Reporting reading 
an updated block would take precedence over reporting ECC 
corrections or retries.  

                     Table xx-xx: Write-Once Parameters

==============================================================================
  Bit |   7    |   6    |   5    |   4    |   3    |   2    |   1   |   0    |
Byte  |        |        |        |        |        |        |       |        |
==============================================================================
 0    |    Reserved     |                   Page Code (TBD)                  |
------|----------------------------------------------------------------------|
 1    |                       Parameter Length  (02h)                        |
------|----------------------------------------------------------------------|
 2    |                                                             | RUBR   |
------|----------------------------------------------------------------------|
 3    |                             Reserved                                 |
==============================================================================

This page may also be applicable for specifying or reporting the 
handling of blocks which might be either blank or written. This 
function is now specified using the Enable Blank Check bit. However, 
there are multiple interpretations that are driven by the differing 
drive implementations. But that's another area altogether.




































@