Date: May 10, 1990 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: May 8-9, 1990 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 am, Tuesday, May 8, 1990. He thanked Peter Blackford of Astro Cable Company for hosting the meeting and he also thanked Nan-Max Travel for arranging the meeting. John also announced that Pat and George Penokie have a new son, Eric Joseph Penokie, born Tuesday, May 1, 1990. John congratulated the new father and thanked Pat for allowing George to travel so soon. 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. The final agenda was as follows: 1. Caching Proposal (90-021R1) [Milligan] 2. Diagnostic Command Set (90-022) [Pickford] {Wednesday a.m.} 3. Packetized SCSI (89-130R1) [Stephens] {Tuesday} 4. Multi-ported SCSI (89-133R1) [Stephens] {Tuesday} 5. SCSI-3 document structure [Lamers] 6. 16/32-bit cable issues (90-48R1) [Penokie] 7. Initiator control of reconnection (90-62) [Lohmeyer] 8. RESERVE command conflict with the direct-access model. (90-059R0) [Penokie] 9. Interaction on LOG SELECT TSD and DS bits (90-067) [Milligan] 10. Report on the Cable Working Group (90-078) [Lamers] 11. Write protect editorial error (90-070) [Snively] 12. Reduced common-mode range on differential SCSI [Jessen] 13. ESDI Editorial Clarifications (90-74) [Lamers] The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Denis D. Springer O 3M Company Mr. Charles Brill P AMP, Inc. Mr. Ed Young O Archive Corp. Mr. Thomas Debiec O Belden Wire & Cable Mr. Thomas Wicklund P Ciprico Inc. Mr. Douglas Hagerman A Digital Equipment Corp. Mr. Rudy Stalzer A Digital Equipment Corp. Mr. Michael Carreira V Digital Equipment Corp. Mr. Fred Zayas V Digital Equipment Corp. Mr. I. Dal Allan P ENDL Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. George Penokie P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. John Lohmeyer P NCR Corp. Mr. John Emerson O Prime Computer Mr. Gene Milligan A Seagate Technology Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Peter Dougherty P UNISYS 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 ------------- -------- ------------- --------------------------------------- X3T9.2/90-59 4/5/90 G. Penokie RESERVE command conflicts with Direct-Access Model X3T9.2/90-62 5/4/90 J. Lohmeyer Initiator Control of Reselection Order Rev 1 X3T9.2/90-67 4/10/90 G. Milligan Interaction on LOG SELECT TSD and DS bits X3T9.2/90-70 5/3/90 R. Snively Editorial Error in SCSI-2 Rev 10C about Write Protect X3T9.2/90-71 5/7/90 T. Wicklund Issues involved with RAID-3 disk arrays using SCSI drives X3T9.2/90-73 5/3/90 R. Snively Proposed Structure for SCSI-3 Documentation X3T9.2/90-74 5/5/90 L. Lamers Request for ESDI Editorial Clarifications Results of meeting 1. Caching Proposal (90-021R1) [Milligan] Gene Milligan's revision 1 proposal was discussed. Several preferences were expressed: * The function of controlling suppression of the MODIFY DATA POINTERS message more properly belongs in the Control Mode page. * Tom Wicklund recommended that more command-by-command caching controls should be provided beyond just the DPO and FUA bits. * Larry Lamers recommended that a method be included to switch the caching algorithm. Regarding the first point above, Dal and Larry suggested that a mode page be defined enable/disable all messages. Larry Lamers objected to the "ascending, contiguous, sequential order" in Forced Sequential Write (FSW) bit description because some defect management schemes may preclude this. It was agreed to drop the "contiguous, sequential order" part. Control of the MODIFY DATA POINTER message will be removed from the FSW bit. The Forced Sequential Read (FSR) bit will be deleted since it is no longer relevant. Gary Stephens asked does FSW imply a write-through? The consensus was that it does not; write-through is controlled by FUA. FSW applies on a per I/O process basis. Gene accepted an action item to revise his proposal and develop another MODE SELECT page or extension to an existing page to enable/disable messages. 2. Diagnostic Command Set (90-022) [Pickford] {Wednesday a.m.} Doug Pickford gave a presentation entitled "Jumbo Shrimp"; which is an oxymoron like "SCSI Diagnostics". Doug is still requesting feedback so that he can develop a new proposal that answers the prior objections while offering the functionality needed to diagnose the underlying hardware. Doug presented the structure for the proposed document and stated that the diagnostics are aimed at rigid disk drives and not at other device types. Dal announced that FlexStar (Mike Witte, VP Marketing, 408-433-0770) has started an industry group that will meet at Spring Comdex to standardized testing of the physical disk drive within SCSI direct-access devices. Phil Devin of DataQuest is also supporting this effort. Doug plans to develop a new proposal for consideration at the next working group meeting. 3. Packetized SCSI (89-130R1) [Stephens] {Tuesday} Discussion on this was brief as Gary has given several previous presentations on packetized SCSI. There were no new questions. The performance improvement potential was extolled by several participants. 4. Multi-ported SCSI (89-133R1) [Stephens] {Tuesday} There was extensive discussion on this topic (about 3 hours). Much time was spent on understanding the concept of grouping. Two terms were defined as follows: A multiple path process allows an initiator to choose one of several paths for an initial connection and the target shall complete the I/O process on the same path. A dynamic path process allows an initiator to choose any path for an initial connection and the target to choose any path at each reconnection for the I/O process. Gary objected to these terms because his proposal is more general. Each port on a device may be given a name. Ports with the same name belong to the same group. I/O processes may use any port with the same group name. Once the concept of completing an I/O process on more than one port is introduced, then it is necessary to add some means to uniquely identify I/O processes across multiple ports. Queue tags are not sufficient. While most people agreed that multiple porting should be added to SCSI-3, many expressed concern that Gary's proposal may go too far. Dal questioned whether the SCSI market really needs or wants the full capabilities (and complexity) associated with dynamic pathing (group size greater than one). IPI has a 40-page working document devoted to this topic. After several hours of discussion, Gary asked if we could now review his document. A number of minor changes were identified and Gary agreed to prepare a revision 2. 5. SCSI-3 document structure [Lamers] Bob Snively gave an introduction to his view of SCSI-3 document structuring (90-073). He identified three documents: SCSI-3E: An evolutionary document containing multi-port, P&L cables, alternate drivers (a small-swing differential), aperture card scanner commands, revised MODE SELECT pages, any new messages needed, the diagnostic command set, and any further ideas that are evolutionary in nature. This document would not replace the SCSI-2 standard; it would contain supplemental information. SCSI-3P: A more revolutionary document which may not be backwards compatible with SCSI-2 or SCSI-3E. It would contain multi- pathing, packetizing, and other revolutionary concepts. SCSI-3FC: This document would contain the necessary information to use SCSI commands over the Fiber Channel. It would include the frame definitions including headers. John Lohmeyer stated that in his opinion these are the possible future markets for SCSI and do necessarily reflect how the documents should be structured. Larry Lamers described his approach to SCSI-3. He is interested in doing an evolutionary document that would supersede the physical sections of SCSI-2, but not affect the later sections. Products complying to the new document would use its physical sections plus the SCSI-2 command sets. He hopes this document could be completed in 12-15 months and expects that it would include the new P/L cables and improved specifications on cable electrical characteristics. It may also include multiple ports, alternative drivers and receivers, and nexus in/out phases. He said he needs more time to think about the best way to structure the other portions of the SCSI-3 standard. John stated that if the committee is in agreement on beginning a SCSI-3 physical layer document then the project proposal should be drafted. 6. 16/32-bit cable issues (90-48R1) [Penokie] Dal Allan reported that Jean Kodama has questioned magnitude of the skew factors on 16- and 32-bit cables. John Lohmeyer said that some NCR chip engineers have also expressed concern over whether mixing 8-bit and 16-bit cables will result in skewed signals due to differences in cable length. John said his (sometimes flawed) intuition on transmission lines is that the cable length difference should not matter as long as the lines are properly terminated. Everyone agreed that testing is necessary. Tom Wicklund volunteered to help, noting that it should be possible to modify the Ciprico SCSI-2 host adapter A/B cable host adapter to use a 16- bit cable. Dal stated that the timing budgets must be defined clearly enough to insure interoperability. Ken Post pointed out that George's document had dropped the requirement to ground reserved lines at the terminators. George will revise his document to say that reserved lines shall be grounded at the terminator and open elsewhere. 7. Initiator control of reconnection (90-62) [Lohmeyer] Tom Wicklund gave a presentation on his paper (90-071) about RAID-3 disk arrays. The presentation supported the NCR request for the messages to provide initiator control of reselection. Tom pointed out that SCSI disk drives are nearly equal in cost to ESDI drives and the array controller interface logic is much less complex (2 chips for SCSI vs. 30 chips for ESDI). However, the integrator does not have control of the read-ahead and reconnection. Control of reselection would dramatically reduce the size of the buffer needed for the RAID controller. Dal noted that the need for initiator control of reselection has come up in the past. Dave Buesing of IBM could probably have avoided his need for a special arbitration algorithm had these messages been available. Tom Wicklund stated that in doing a fiber channel physical layer these messages would be useful in managing the devices attached to the fiber channel controller. The DO NOT CALL ME (a.k.a. INITIATOR RECONNECTION REQUEST) was renamed to DISABLE TARGET RECONNECTION. After some additional discussion, Bob Snively suggested that it might be better to only disable target reconnections for data transfers. This would permit a target to reconnect and immediately disconnect when the data is ready for transfer. John Lohmeyer accepted an action item to revise his proposal for the next meeting. 8. RESERVE command conflict with the direct-access model. (90-059R0) [Penokie] John reviewed the conflict and reported on his research into prior minutes. The minutes said that reservations and prevention of medium removal are independent concepts. However, the minutes shed no additional light on what was really meant by "independent". Larry continued to maintain that independent only meant that prevention of medium removal does not imply an automatic reservation. PREVENT ALLOW MEDIUM REMOVAL commands with a prevent bit of zero are allowed through reservations to avoid deadlock conditions. If such commands with a prevent bit of one were also allowed through reservations, then deadlocks could occur. Paul Boulay had registered his opinion on the subject via a message on the SCSI BBS which John read. He supported Larry's position and George Penokie also joined their position. Those opposing this position mostly disliked the fact that one must check the prevent bit before deciding whether a PREVENT ALLOW MEDIUM REMOVAL command may be accepted or rejected when a reservation is active. They interpreted "independent" as meaning both flavors of the command should be executed in spite of a reservation to another device. The new fathers gradually won over most of the opposition and Larry accepted an action item to draft wording for a motion at the next plenary meeting. 9. Interaction on LOG SELECT TSD and DS bits (90-067) [Milligan] Gene Milligan accepted an action item to develop a motion that will resolve the conflict in LOG SELECT and LOG SENSE through selective surgery. That is, excising the last two sentences of paragraph 13 of section 7.3.2. 10. Report on the Cable Working Group (90-078) [Lamers] See X3T9.2/90-078 for minutes of the meeting. In a nutshell, several cable vendors have completed the initial testing, the cross talk measurements require correlation which Jim Fiala of 3M is undertaking, a second iteration on the cable testing will be done based on the initial results, and the next cable meeting is set for Tuesday, June 19, 1990 at 1:00 p.m. An ad-hoc meeting on connectivity issues is planned for Tuesday, June 19, 1990, at 10:00 a.m. The purpose of this meeting is to address the issue of how to use high-density connectors in assembling systems. Connector, cable, cable assembly, and system vendors are invited to demonstrate their solutions. 11. Write protect editorial error (90-070) [Snively] The definition of the write protect (WP) bit in the device specific parameter byte in MODE SENSE data is not consistent in sections 8, 9, and 15. There was a short discussion, however prior to a final agreement several members need to consult with the folks back home. Based on the original wording in SCSI-1 the WP bit is reserved in the MODE SELECT data and an indicator of the write protect condition in MODE SENSE. Larry Lamers accepted an action item to draft wording to sychronize the wording of the WP bit in sections 8, 9, and 15. The wording will be based on the input from HP, Seagate, Sun, and Maxtor. The draft wording is in preparation for a motion at the next plenary meeting. 12. Reduced common-mode range on differential SCSI [Jessen] Erik Jessen questioned why SCSI-2 specifies +12/-7 volt common-mode range on differential drivers/receivers. If SCSI really experiences its specified range significant currents would flow through the ground wires. He pointed out that HiPPI uses a 1 volt range on 10 mA ECL drivers. However, HiPPI is only point to point. IPI has 25Mhz ECL but it is not multi-drop, only point to point. There were hopes that reducing this range and reducing the voltage swing could result in a differential drive option which could be integrated into one or two chips. Eric Jessen accepted an action item to investigate reduced differential common-mode range and the cable requirements. 13. ESDI Editorial Clarifications (90-74) [Lamers] Larry said that Maxtor had found several editorial issues with ESDI. He had discussed these with Dal and Tom Wicklund. He plans to review the issues with other ESDI-interested parties prior to the June plenary meeting and expects to make a motion directing the technical editor to make some editorial clarifications during the final editing. Review of action items 1) Larry Lamers will draft wording to sychronize the wording of the WP bit in sections 8, 9, and 15. 2) Larry Lamers will draft wording for resolving the RESERVE command conflict with the direct-access model. 3) John Lohmeyer will revise his proposal for initiator control of reconnection. 4) Eric Jessen will investigate reduced differential common-mode range and the cable requirements. 5) Doug Pickford will develop a new proposal on diagnostics for consideration at the next working group meeting. 6) Gary Stephens will revise the multi-ported SCSI proposal to incorporate the results of the discussion at May working group meeting.