Date: March 8, 1990 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: March 6-7, 1990 X3T9.2 Working Group Meeting John Lohmeyer opened the meeting by thanking Jeff Stai of Western Digital for hosting the meeting in Costa Mesa, CA. John noted that the Red Lion Inn had excellent meeting facilities and the phones included a data port for modems for the Power Users who would rather stay in their hotel rooms and play with telephones and computers instead of fighting Orange Co. trafffic. John also thanked Larry Lamers and Maxtor for providing draft copies of SCSI-2 Rev 10c for review at the meeting. The final agenda was as follows: 1. Proposed REQUEST TEXT SENSE Command (89-160) 2. Dealing with queued commands during immediate conditions [Penokie] 3. Adding Unrecorded Blocks in Buffer Count field [Racelo] 4. Review SCSI-2 rev 10c 7.3.2 Log Parameters 8.2.11 Paragraph 1 RELEASE command 8.2.11.1 Third-Party clarification 8.2.12.1 Paragraph 3 8.2.4 Paragraph 4 PREVENT ALLOW MEDIUM REMOVAL failed cache sync. 5. Caching Proposal (90-021) [Milligan] 6. Diagnostic Command Set [90-022] [Pickford] 7. Differential Transceivers [Murdock] 8. Packetized SCSI [Stephens] 9. Multi-ported SCSI [Stephens] 10. SCSI-3 document structure [Lamers] 11. 16/32-bit cable issues [Penokie] The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Ed Young O Archive Corp. Mr. Bharat Shah P Cipher Data Products, Inc. Mr. Mike Racelo P DDC Pertec Mr. Paul Hanmann P Emulex Corp. Mr. Chuck Micalizzi A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Denni Rash V FileNet Corp. Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. George Penokie P IBM Corp. Mr. Paul Anderson A IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. John Lohmeyer P NCR Corp. Mr. Dennis P. Trupski P Olivetti Mr. Gene Milligan O Seagate Technology Mr. Fred Burgess S Seagate Technology Ms. Rita Lin P Sony Corp. of America Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Ricardo Dominguez S Texas Instruments Mr. Peter Dougherty P UNISYS Mr. James Patton P Visqus Mr. Doug Pickford A Western Digital Mr. Erik Jessen O Western Digital 25 People Present Status Key: P Principal A Alternate O Observer S Special Interest (frequent visitor) V Visitor The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/86-109 3/2/90 Too Many Draft of extended Small Computer System Rev 10c Interface, "SCSI-2" X3T9.2/90-21 3/6/90 G. Milligan Additional SCSI Caching Control Rev 1 X3T9.2/90-28 3/4/90 M. Racelo Modifying Actual Retry Count/Adding Rev 1 Unrecorded Block Count in Buffer X3T9.2/90-46 3/5/90 B. Shah Reporting Number of Blocks/Bytes in Target's Buffer X3T9.2/90-47 3/5/90 G. Stephens Modifying Actual Retry Count X3T9.2/90-48 5/2/90 G. Penokie 16/32 bit P/Q and L cable stand alone document Results of meeting 1. Proposed REQUEST TEXT SENSE Command (89-160) [Cawley] John Lohmeyer read a letter from Alec Cawley of Quantel stating the need for a REQUEST TEXT SENSE command. This command would return normal REQUEST SENSE data plus the ASC/ASCQ description in ASCII. There was initial enthusiasm for this idea until Gene Milligan asked about the need for a field to identify which language is in use for the text data. Multiple language support quickly soured the whole idea. Peter Dougherty mentioned that language support is usually handled by the operating system via special console drivers. The sense codes are interpreted and the appropriate text in the desired language is displayed on the console. Doing the translation in a peripheral would add complexity. They probably would not use the REQUEST TEXT SENSE command even if it were supported. The consensus of the working group was to abandon the proposal. John Lohmeyer said he would write a letter to Alec Cawley explaining the decision. 2. Dealing with queued commands during immediate conditions [Penokie] How does the target deal with other initiators during immediate operations. Is the device 'NOT READY'? Does the command get queued? What happens when the immediate operation completes? George Penokie asked the question of what should a target do if it is executing a FORMAT UNIT command with an immediate bit of 1 and it receives another command from another initiator. George argued that this should be treated the same as if the FORMAT UNIT command had an immediate bit of 0. George had interpreted the standard as permitting the target to queue such commands. Almost no one else agreed with George's interpretation. After considerable discussion, the group reached consensus that while a FORMAT UNIT is in progress, regardless of the state of the immediate bit, the medium is logically unable to accept a medium-access command. Therefore the unit is not ready and the target should report NOT READY sense key. In order to make this clearer, the description of the target response was moved from the paragraph on the immediate bit to section 8.2.1. This wording will also be clarified so that it does not conflict with the requirement to return RESERVATION CONFLICT if a reservation is active. 3. Adding Unrecorded Blocks in Buffer Count field [Racelo] Several people with interest in sequential-access devices met at lunch to discuss Mike Racelo's document and the responses from Gary Stephens (90-47) and Bharat Shah (90-46). The conclusion was that the information that Mike requested be added is contained in the READ POSITION command except for the number of filemarks and setmarks. No requirement for this information has been identified. No changes will be made until a need for this information is identified. 4. Review SCSI-2 rev 10c Larry Lamers lead a discussion of all the editorial changes made to SCSI-2 since Rev. 10b. A number of further editorial improvements were identified during the meeting. SCSI-2 Rev. 10c will be included in the March mailing and a motion to substitute Rev. 10c for 10b is expected at the April plenary meeting. The editorial changes (other than minor grammar) are: 1. Several minor inconsistencies between command names in the body of the document and those used in Appendix I have been corrected. 2. Three missing operations codes were added to the summary table in Appendix I. 3. Table headings were added to the second and subsequent pages of the multi-page tables. 4. Notes were added to the operation code and ASC/ASCQ tables to point out that these tables are present in both alphabetical and numeric order. 5. A note was added in three places to alert readers that X3T9.2 is considering an alternative wide data transfer method employing a single 16-bit cable. 6. The cable specifications were clarified to show that 28 AWG wire is only required on lines that distribute power. Signal lines may use smaller gauge wires. While cable impedances above 90 ohms are recommended, those with impedances below 90 ohms are now permitted since such cables have been demonstrated to operate successfully and they are more readily available. 7. The document was ambiguous on whether a transfer period of 200 ns should use the new fast data transfer timings. This was clarified so that 200 ns transfer periods use the slow timings, which is compatible with SCSI-1. 8. A subsection on "Unexpected Reselection" was added to recommend that the initiator send an ABORT message to a target that attempts to reselect an initiator when there is no valid nexus. 9. The document was clarified so that the requirement that a target not accept both tagged and untagged queued commands applies on a per- initiator basis. It would be difficult to manage a multi-initiator system if all initiators were required to use the same queuing method. 10. The IT8 Committee's peripheral device type name was corrected to "Graphics Arts Pre-Press Devices". 11. An implementors note was added to the mode pages to suggest a method to improve compatibility with SCSI-1 initiators. 12. An implementors note was added to the CHANGE DEFINITION command to list possible areas where SCSI-2 target behavior may cause compatibility problems with SCSI-1 initiators. 13. Several omitted ASC/ASCQ codes were added to the ASC/ASCQ table. 14. The RELEASE command description was clarified to avoid misinterpretations regarding the release of third-party reservations via a non-third-party RELEASE command. Third-party reservations are only released by third-party RELEASE commands. The requirement to "ignore" invalid attempts to release a reservation was moved from the RESERVE command description to the RELEASE command description, where it properly belongs. Similar changes were made to the RESERVE UNIT and RELEASE UNIT commands in section 9 and to the RESERVE and RELEASE commands in section 16. 15. The list of commands that may be accepted while a unit reservation is active was updated to include a RESERVE command in section 8 and a RESERVE UNIT command in section 9. Section 16 already included the RESERVE command in the list. This change is necessary because of the requirement to accept superseding reservations. 16. An implementors note was added clarifying the target actions when a write command encounters the physical end-of-medium. This note also explains the different interpretation which was possible in SCSI-1. 17. The models for the optical memory devices now include a statement that media standards may impose other requirements regarding defect management. 18. "Power-up" was changed to "power-on" for consistency. 19. The Vendor ID code table in Appendix J was updated to include recent additions. 20. An implementors note was added to section 7.2.10.5 recommending a method for SCSI-2 target's to improve their compatibility with SCSI-1 initiators. 21. A section (8.1.10) was added to the direct-access device model to mention rotational position locking. This concept was missing from the model, but no new requirements were added. 22. The direct-access device model now has statement explaining the defect management differences between fixed and removable media. 23. Definitions of "buffered mode" and "spacing" were added to the sequential-access device glossary. 24. The READ BLOCK LIMITS command description was clarified. 25. The last sentence of the seventh paragraph of section 8.3.3.6 was corrected to "The TB bit does not affect the action taken for recovered data." 26. The implementors note in 5.1.3 was expanded to document that SCSI-2 devices are permitted to accept the SCSI-1 select-without-ATN protocol as documented in SCSI-1. 27. Table 7-7 was merged with the following in-line table and the terminology was clarified. Previously, the term "direct-access" was used for device types 0, 4, 5, and 7. But device type 0 is known as a "direct-access device". The term "block device" is now used for this group. A similar conflict existed for sequential-access devices (device types 1, 2, 3, and 9). This group is now identified as "stream devices". The paragraph permitting device type 5 to be a destination in the case of a COMPARE command was incorporated as a note in Table 7-7. 28. The description of the VFU Control Byte in the Parallel Printer Interface Parameters mode page was made into an implementors note since it contains no requirements and the VFU Control Byte is not part of the SCSI-2 standard. 5. Caching Proposal (90-021) [Milligan] Gene Milligan presented several foils on his caching controls proposal. He received several suggestions concerning what is meant by "zero latency". Gene revised his proposal over night and provided copies of revision 1 before the end of the meeting. 6. Diagnostic Command Set [90-022] [Pickford] Doug presented slides trying to build support for creating a Diagnostic Command Set (DCS) document and having a "diagnostic mode" whereby hosts can access hidden functions in the device. Most of the drive vendors were opposed to standardizing these functions even though most include vendor-specific diagnostic functions in their products. Some of their concerns include fears about the security of their proprietary microcode, exposing the low-level functions of their drives to potential corruption from end-users, and concerns about the standard restricting future enhancements. The systems integrators who were present were divided on the need for a standard DCS. Bob Snively was opposed to several of the proposed functions, especially having both a DATA IN and a DATA OUT phase in the same I/O process. Bob was also opposed to the notion of a "diagnostic mode". He preferred to merely add diagnostic commands without any special states or modes. Much of the morning was spent with Dal shouting at Bob Snively and Bob shouting back. Dal said he just wanted Doug to get a fair hearing and that the people who want a standard DCS were under-represented at the meeting. It may be difficult to achieve consensus over DCS. However, not all comments were negative. The discussion ended on a positive note with the following suggestions to Doug: 1) He should revise his proposal to use the SEND DIAGNOSTIC and RECEIVE DIAGNOSTIC RESULTS commands instead of using both DATA IN and DATA OUT phases in the same command. Expanding the CDB length also will result in compatibility issues with present chips and host adapters. 2) The commands should be more generic. 3) A method should be provided for security of the microcode and to prevent unauthorized access to the diagnostic functions. Doug plans to revise his proposal for the March mailing. It will be on the agenda for the May working group meeting. 7. Differential Transceivers [Murdock] Gary was not present so this item was deferred until May. 8. Packetized SCSI [Stephens] Gary was not present so this item was deferred until May. 9. Multi-ported SCSI [Stephens] Gary was not present so this item was deferred until May. 10. SCSI-3 document structure [Lamers] Larry requested that this item be deferred until the April plenary meeting. 11. 16/32-bit cable issues [Penokie] George presented 90-48 document which consolidates his previous proposals. All previous proposals are now obsolete. 90-48 includes examples of A to P, A to L, and P to L cable converters. This issue of terminator power gets much worse when considering more than 8 devices and wide cables. It may be necessary to provide terminator power external to the SCSI cable. This topic will be included on the May working group agenda. George had also added a field to INQUIRY which specifies which wide options are supported.