Date: January 16, 1991 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: January 15-16, 1991 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 a.m., Tuesday, January 15, 1991. He thanked Ken Post of Future Domain for hosting and arranging the meeting. As is customary, the people attending introduced themselves. A copy of the of the X3T9.2 membership list was circulated for attendance and corrections. Copies of the draft agenda and the recent document register were made available to those attending. Information on X3T9.2 and Mailing Subscription Forms were made available. John Lohmeyer reviewed the current status of SCSI-2. Following the plenary vote to withdraw SCSI-2, the full complexity of the process became known. X3T9 did not act on the request because it would require a letter ballot. Del Shoemaker has negotiated an improved process not only for ESDI and SCSI-2 but also future standards. While not finalized, the proposed process would have the ANSI editing done in parallel with public review, such that a clean document goes to X3 for final balloting. We also hope to get the ISO editing review done at the same time. John Lohmeyer reported that the ESDI document will be sent out for a one-month public review due to changes during final edit. He did not know when this review will occur. ANSI and CBEMA have agreed to return SCSI-2 to the committee for editing as if the committee had agreed to make substantive changes following the first public review period. The revised document will require a two-month public review period. John gave an overview of the Low-Power Differential working group meeting which was held Monday afternoon (see 91-7). Mike Bartlett of T.I. had been absent so much of the planned agenda was not covered. The final agenda was as follows: 1. Long BUSY conditions (90-168) [Milligan] 2. Additional LUNs (90-183) [Penokie] 3. SCSI-3 Packetized Protocol (SCFI) (90-132R3) [Stephens] 4. Proposed change to bus driving rules in 90-048 R7 (91-1) [Lohmeyer] 5. SCSI-2 Technical Editing Questions [Lamers] 6. IEEE P-1394 discussion {Tuesday 1:00 p.m.} 7. New ASCQs (90-184) [Penokie] 8. SCSI-3 RESERVED lines (90-182) [Lincer] 9. ATA issues {Tuesday 4:00 p.m.} 10. Signal Quality issues for SCSI-3 11. Diagnostic Command Set (90-103R2) [Pickford] {Wednesday 10:00 a.m.} 12. P Cable Retention mechanism (90-187) [Chan] 13. Concurrent execution of more than one I/O Process [Pepper, Penokie] {Wednesday p.m.} 14. CD-ROM Starting Logical Block Address (90-200,-202) [Simpson, Boulay] 15. Non-ASCII Characters in Log List Parameters (90-201) [Penokie] 16. ESDI Power Pin Diameter Issue (90-199) [McGrath] 17. Analysis of long single-ended buses (91-2) [Chan] 18. Changing the notch page () [Pepper] 19. Review of Dual port document [Houlder] 20. Issue on AEN in Appendix F [Houlder] 21. Reselection and ATN - Stai - Stephens The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/90-136 1/8/91 G. Houlder Extentions for dual port SCSI Rev 3 X3T9.2/90-183 11/15/90 G. Penokie Increasing LUN Addressability X3T9.2/90-184 11/16/90 G. Penokie New Additional Sense Code Qualifiers X3T9.2/90-201 12/20/90 G. Penokie Non-ASCII Characters in Log List Parameters X3T9.2/91-1 1/15/91 J. Lohmeyer SCSI-3 Wide Bus Driving Rules Selection Rev 1 and Reselection Phases X3T9.2/91-2 1/13/91 K. Chan Analysis of long single-ended buses X3T9.2/91-3 1/14/91 M. Teener P1394 High Speed Serial Bus -- A Technical Summary X3T9.2/91-5 1/15/91 S. Anderson Addition to Set Features command X3T9.2/91-6 1/16/91 M. Peper How does one change the Active Notch field in the Notch Mode page? The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Robert C. Herron A 3M Company Mr. Thomas Newman S Adaptec, Inc. Mr. Charles Brill P AMP, Inc. Mr. Jon Abilay A Apple Computer Mr. Michael Teener V Apple Computer Mr. Ed Young P Archive Corp. Mr. Fernando Ramirez O Burndy Mr. Edward Hrvatin O Burndy Corp. Mr. Bharat Shah A Cipher Data Products, Inc. Mr. Robert Kellert P Cirrus Logic Inc. Mr. Steven Anderson S Conner Peripherals Mr. Douglas Hagerman A Digital Equipment Corp. Mr. John A. Gallant S Digital Equipment Corp. Mr. Paul Hanmann P Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. Howard Wang O Hitachi Mr. Mario Yamaguchi V IBM (Japan) Mr. George Penokie P IBM Corp. Mr. Gerald Marazas A IBM Corp. Mr. Paul Anderson A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Gary R. Stephens S IBM Corp. Mr. Gregory Floryance S IBM Corp. Mr. Rodrigo Samper V IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Gary Murdock A National Semiconductor Mr. Thai Nguyen S National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. Steve Kohlenberger O Presoft Architects Mr. James McGrath P Quantum Corp. Mr. Gene Milligan A Seagate Technology Mr. Gerald Houlder A Seagate Technology Mr. Brian Johnson O Seagate Technology Mr. Fred Burgess S Seagate Technology Mr. Robert L. Simpson P Sony Corp. of America Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Arlan P. Stone A UNISYS Mr. Jeff Stai P Western Digital 43 People Present Status Key: P - Principal A - Alternate O - Observer S,V - Visitor RESULTS OF MEETING 1. Long BUSY conditions (90-168) [Milligan] Jim McGrath proposed adding a time field (a la the AEN Holdoff Period). This field would define the maximum time the target could delay (preferably disconnected) before it must report BUSY status or CHECK CONDITION status, depending on the RAC bit. The group agreed on a two-byte field that would use increments of 100 milliseconds. A value of zero would be undefined (as is currently the case). A value of FFh would mean there is no upper limit. The group considered two possible sense keys: ABORTED COMMAND or NOT READY sense key with an appropriate ASC. NOT READY was favored by most people, but Gary Stephens pointed out that some sequential-access drivers, especially those that do not inspect ASCs, might confuse this situation and request operator assistance in making the unit ready. Gene Milligan read the revised wording developed the previous evening. The discussion went off into obtuse corners. Larry Lamers suggested that a changing a "shall" to "should" in SCSI-2 Rev 10d. This would relax the requirement that BUSY status be returned during ECA situations and would make it legal to return other statuses during ECAs. No consensus was reached. 2. Additional LUNs (90-183) [Penokie] While some members of the group did not agree with George's example for the need of more logical units, there was consensus that the field in the IDENTIFY message for Logical Unit Number should be expanded to 5 bits in SCSI-3. The working group recommends that the plenary group accept this proposal for SCSI-3. 3. SCSI-3 Packetized Protocol (SCFI) (90-132R3) [Stephens] The working group suggested that Gary remove considerable material from 90- 132R3 that covers concepts outside the scope of the SCSI-3 Packetized Protocol. This material should be considered for other SCSI-3 documents. Gary plans to prepare a lower-fiber content (that is, fewer pages) document. 4. Proposed change to bus driving rules in 90-048 R7 (91-1) [Lohmeyer] John Lohmeyer summarized NCR's suggested changes to accommodate multiple- chip 32-bit wide implementations. Any byte that has a bit set and all lower bytes are required to be driven during SELECTION and RESELECTION phases. Only these bytes are checked for correct parity. Gerry Houlder pointed out a typo which John corrected in revision 1 of his document. The working group recommends that the plenary group accept this proposal for SCSI-3. 5. SCSI-2 Technical Editing Questions [Lamers] Larry Lamers reported that several people had met to draft a new paragraph for the CD-ROM section clarifying the location of logical block address zero. Document 90-200R1 contains the proposed paragraph. It is on the agenda for the February meeting. Paul Boulay investigating reports from ANSI that a table is missing in section 16. Gary Murdock pointed out that ISO 8482 is not entirely the same as RS-485 concerning the required voltage range that the part is supposed to tolerate. The group asked Gary to provide proposed wording to correct the SCSI-2 document. Larry Lamers reported that Mr. Mori, a Japanese expert from SC 25/WG 4, had asked why Table 7-3 (CHANGE DEFINITION) does not have code values to specify ISO SCSI and ECMA SCSI. Several opinions surfaced. The majority of the working group felt that CHANGE DEFINITION command in the ANSI document should not be expanded because there is almost no difference between ANSI SCSI-1 and ISO SCSI-1 (no one knew of any implementations of the one permissible difference). The CCS code value refers to a non-ANSI, non-ISO document. At this point in time, there is no technical difference in the ANSI and ISO SCSI-2 documents. This decision may need to be re-addressed in the ISO document if it is changed significantly from the ANSI document. 6. IEEE P-1394 discussion {Tuesday 1:00 p.m.} Michael Teener, Chair of IEEE P1394, gave a presentation on the P1394 High Speed Serial Bus (91-3). The goal is a low cost (<$15) peripheral bus that bridges backplane buses and supports peripheral devices. The maximum distance is 10 meters. The current specification uses a 50 Mbaud/sec signaling rate with the FDDI 4B/5B encoding to achieve burst speeds of 40 Mbit/sec. A lot of protocol has been borrowed from FDDI including the FDDI-II concept of isochronous channels. The topology is quite flexible and the network self-configures as devices are powered on. The principal application intended for P1394 is to provide a standard interface for such devices as printers, keyboards, mice (pointing devices), etc. However, there is sufficient bandwidth to consider using this interface on some storage devices. Michael invited interested parties to attend the next P1394 meeting on 1/25/91 in Santa Clara, CA. Michael can be reached at 408-974-3521. He has an email address of: Teener@Apple.com. 7. New ASCQs (90-184) [Penokie] After some discussion, the working group decided to recommend that the new ASCQs be accepted for SCSI-3. Gene Milligan did not think the list was complete, however he had no further additions at this time. 8. SCSI-3 RESERVED lines (90-182) [Lincer] Dave Lincer brought this item up at the San Jose meeting. The issue was whether targets that include optional termination should provide independent switching of the ground line to the RESERVED lines or could all the RESERVED lines be connected together. Unfortunately, the answer to this question very much depends on what will be done with the RESERVED lines. There is no certain answer at this time. It was pointed out that the Small Form Factor industry group expects to use these lines for power (+5 V). John Lohmeyer said he had received several inquiries from people who would like to use the Terminator Power lines for more than powering the terminators. While it seems likely that some of these lines will eventually be used for power, it still is uncertain whether all of the lines will be used for this purpose. 9. ATA issues {Tuesday 4:00 p.m.} Steve Anderson of Connor presented a proposal (91-5) to add a 66h code to the SET FEATURES command to prevent resetting of features after a software reset. A CCh code will disable this feature. The group agreed with the proposal and Dal plans to include the proposal in the next revision of ATA. 10. Signal Quality issues for SCSI-3 Kurt Chan presented an analysis of the theory behind Bill Spence's empirical observations on cable signal quality using a Bergeron diagram. Cable specifications on attenuation depend on the frequency measured. It is very important that the cable is measured in the manner in which it is used. Many single ended cables 25 meters in length have 25% attenuation and thus require 2.6 volts input to get 2.0 volts at receiving end on first step. 80 ohm cables provide nearly equal amounts of low and high margin for 25 meter cables using 110 ohm termination and thus are optimal for these lengths with typical attenuations. This phenomenon is due to cable attentuation, not cable impedance. If 6 meter cables are considered, the balanced impedance value changes. The SCSI-2 specifications are appropriate at this length if the signal pairs are properly placed. Kurt recommended: 1) We need a uniform method of measuring attentuation. 2) People should use 64 mA drivers, to get the near end voltage as low as possible. Bill Spence presented his results using a 68-foot cable. A discontinuity was exhibited on two of the three drives tested. Apparently when the chip switches a different impedance is reflected back. Higher impedance cables exacerbate this phenomena. Bill also showed a synchronous transmission in which every other assertion was not as good. This could be a cable charging effect due to the close spacing of two bytes, as might happen when a host adapter assembles two bytes into a word for transfer on a backplane. This phenomenon could have serious effects on a 32-bit wide backplane using fast SCSI. 11. Diagnostic Command Set (90-103R2) [Pickford] {Wednesday 10:00 a.m.} Doug Pickford was unable to attend the meeting so the review of his document was deferred until the March working group meeting. Dal Allan requested that Doug's critics "bring paper" documenting their issues and/or offering alternative proposals. 12. P Cable Retention mechanism (90-187) [Chan] The working group had no objection to the idea of specifying a 2-56 jackscrew made from high tensile-strength steel for the SCSI-3 P Cable connector. A motion to this effect was planned for the February plenary meeting. Discussion on this item was set for 4:00 p.m. Monday afternoon. The X3T9.3 who care about this issue will be invited to participate. 13. Concurrent execution of more than one I/O Process [Pepper, Penokie] George Penokie presented a diagram of the issues. George's explanation interprets 'active I/O process' as movement of data to/from the medium, and 'current I/O process' as activity on the SCSI bus. Thus two different I/O processes, one active and one current, could be occurring simultaneously. A lengthy discussion ensued in which it was evident that several interpretations of the terminology were in use. Additional definitions may be needed. There was general agreement that multi-tasking was desirable, and the difficulty was in getting agreement on how to describe it. 14. CD-ROM Starting Logical Block Address (90-200,-202) [Simpson, Boulay] This item was covered during the discussion of agenda item 5. 15. Non-ASCII Characters in Log List Parameters (90-201) [Penokie] George's proposal would add a bit in SCSI-3 to allow LOG SENSE to return binary data in list parameters. George agreed to prepare a revision of his document calling the data 'binary' instead of 'hex'. The working group recommends that the plenary group accept this proposal for SCSI-3. 16. ESDI Power Pin Diameter Issue (90-199) [McGrath] Jim McGrath of Molex has found an error in the ESDI document where the diameter of the pins in the power connector are not specified. John explained the situation and stated that this would be resolved at the plenary. Dal Allan, the project editor, agreed to contact the ANSI editor to see if this can be fixed before the proposed one-month public review on the other ESDI final editing changes. 17. Analysis of long single-ended buses (91-2) [Chan] This agenda item was covered under the related agenda item 10. 18. Changing the notch page (91-006 R0) [Peper] Mike Peper found a logical conflict in the SCSI-2 standard concerning changing the active notch in the Notch and Partition Page. It is not logically possible to switch notches without knowing in advance the information about the starting and ending boundaries. These fields are not changeable, but the initiator is required to reflect their values from the MODE SENSE values. Targets are required to return CHECK CONDITION status and ILLEGAL REQUEST sense key if the initiator guesses wrong about these fields. Unfortunately, the initiator cannot obtain their values without switching the active notch. Catch 22. Mike's solution is to add wording to SCSI-2 that requires the target to ignore the values in these fields for the MODE SELECT command. The working group recommends that this correction be accepted for SCSI-2. 19. Review of Dual port document (90-136 R3) [Houlder] Gerry reviewed his latest revision. This document reflects the comments made at the last plenary meeting. Dual porting is now only defined for two ports, but the architecture can be extended to more than two ports. The effect of resetting one of the ports is defined and and several other minor items were clarified. See document 90-136 Revision 3 for details. There was some discussion on the re-mapping of the four bits in the INQUIRY data into an enable bit and 1-bit port number (extensible to 3 bits). This is not consistent with the vendor-specific approach that one company that has already done. Since that company would need to make minor firmware changes in any case, the group opted for the more extensible solution. Gerry was requested to add in a statement clarifying that there is no logical association between devices on the two cables. Thus a host that is capable of being an initiator on both ports cannot obtain REQUEST SENSE data for port A from port B of a dual-ported device -- the ports are logically separate devices (even though they both access the same medium). Consecutive READ commands on the two ports to the same logical block address on a direct-access device would return the same data. 20. Issue on AEN in Appendix F [Johnson] Brian Johnson pointed out that appendix F suggests that most devices do an INQUIRY command followed by the verify state test. Appendix F.5 suggests the opposite approach. Bob Snively, who drafted this appendix, said there was no reason for the reversal and that the INQUIRY command should be sufficient. Dal Allan and Larry Lamers suggested better ways of dealing with the problem of AEN initialization. Their recommendation is wait until an asynchronous event occurs, then do INQUIRY commands to all the devices that have issued commands to the target since power on. This procedure defers the AEN survey until the last possible moment and repeats it on every AEN. No devices (which may have changed the state of their AENC bit) are missed. The working group did not draft a concrete recommendation on what should be done (if anything) about section F.5. 21. Issue on ATN & IDENTIFY & RESELECTION - [Houlder, Stephens, & Stai] This issue involves what to do about various error conditions that might arise during a reconnection prior to fully identifying the I/O Process. There are some ambiguous circumstances: 1) What should a target do if ATN is true during RESELECTION phase? Section 5.2.1 requires that the target transfer the first MESSAGE IN byte (IDENTIFY) prior to honoring the ATN signal, but there may also be a SIMPLE QUEUE TAG message pending. Should 5.2.1 also require transferring the SIMPLE QUEUE TAG message? 2) Jeff Stai had pointed out that the IDENTIFY message implies that the implicit RESTORE POINTERS function should be performed by the initiator possibly before the I/O Process is fully identified. That is, before the SIMPLE QUEUE TAG message has been successfully transferred. Gary Stephens objected to the special ATN handling specified in 5.2.1, item (6). After a lengthy discussion, the chair requested a document that proposes specific wording changes. Gary Stephens agreed to draft a proposal for the February plenary meeting.