Date: July 11, 1990 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: July 9-10, 1990 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 am, Tuesday, July 10, 1990. He thanked George Penokie, Horst Truestedt, and Dave Buesing of IBM 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. The final agenda was as follows: 1. Caching Proposal (90-021R2) [Milligan] 2. Diagnostic Command Set (90-103) [Pickford] {Wednesday a.m.} 3. Proposed requirement for Diagnostic Command Set (90-89) [Snively] 4. Packetized SCSI (89-130R1) [Stephens] 5. Multi-ported SCSI (89-133R1) [Stephens] 6. 16/32-bit cable issues (90-48R1) [Penokie] 7. Can TDRD be used to block data transfer on current connection? (90- 062R3) [Lohmeyer] 8. SCSI-2 Request for Interpretation #1 (90-98) [Earle] {Should there be a Page Size Descriptor for each partition on the tape?} 9. SCSI-2 Request for Interpretation #2 (90-97) [Earle] {Should filemarks/setmarks be assigned block addresses?} 10. Reduced Common-Mode Range on Differential SCSI (90-092) [Jessen] 11. SCSI-2 Alt-2 Termination for the B Cable [Wicklund] 12. SCSI-3 Physical Project Proposal [Lohmeyer] 13. Defect list length exceeding 64kb (90-110) [Penokie] 14. Request for clarification during logging operations (90-111) [Penokie] 15. Editorial clarification in SCSI-2 [Penokie] 16. High-density connector specification issues (90-107) [Chan] 17. 1991 Meeting planning [Lohmeyer] The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Patrick E. Pejack S Adaptec, Inc. Mr. Jon Abilay A Apple Computer Mr. Ed Young P Archive Corp. Mr. Bharat Shah A Cipher Data Products, Inc. Mr. Douglas Hagerman A Digital Equipment Corp. Mr. Chuck Micalizzi A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Terry Maezawa P Furukawa Electric Amer, Inc. Mr. Kenneth Post P Future Domain Mr. George Penokie P IBM Corp. Mr. Gerald Marazas A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. John Lohmeyer P NCR Corp. Mr. David Steele S NCR Corp. Mr. Dennis P. Trupski P Olivetti Mr. Gerald Houlder A Seagate Technology Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Edward R. Schurig O Texas Instruments Mr. Ricardo Dominguez S Texas Instruments Mr. Doug Pickford A Western Digital Mr. Erik Jessen O Western Digital 24 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 ------------- -------- --------------- --------------------------------------- X3T9.2/90-62 7/5/90 J. Lohmeyer Initiator Control of Reselection Order Rev 3 X3T9.2/90-89 6/15/90 B. Snively Proposed requirement for Diagnostic Command Set X3T9.2/90-92 6/17/90 E. Jessen Alternative Differential Standard and its impact X3T9.2/90-97 6/14/90 B. Earle Req. for Interpret. (#2) SCSI-2 Seq-access Recorded Obj. Identifiers X3T9.2/90-98 6/14/90 B. Earle Req. for Interpret. (#1) SCSI-2 Seq-access Partition Descriptors X3T9.2/90-100 6/26/90 B. Spence SCSI Cables Observations and Proposals X3T9.2/90-103 7/2/90 D. Pickford Proposed Extensions for SCSI-3, Standard Physical Layer Access X3T9.2/90-105 8/20/90 J. Lohmeyer Draft Project Proposal for SCSI-3 Physical Level Standard X3T9.2/90-107 7/6/90 K. Chan High Density Connector Specification Issues X3T9.2/90-109 6/28/90 G. Stephens More SCSI Tape Stuff X3T9.2/90-110 7/3/90 G. Penokie Defect List Lengths exceeding 64 Kbytes X3T9.2/90-111 7/5/90 G. Penokie Request for Clarification on exception conditions during log operations X3T9.2/90-112 7/6/90 B. Shah Interpretation of LOCATE and READ POSITION Commands X3T9.2/90-113 7/6/90 B. Shah Interpretation of Mode Parameters Medium Partition Page (1) RESULTS OF MEETING 1. Caching Proposal (90-021R1) [Milligan] Gerry Houlder distributed copies of a memo from Gerry to Gene Milligan proposing that several modifications be made to Gene's caching proposal. The working group discussed Gerry's proposed changes. Larry Lamers suggested that the ACDI field (a field to control how quickly an adaptive cache would adjust itself) not be tied into the number of I/O processes. He suggested a more generic control mechanism: 0 = off, 127 = on at nominal level, 1 is adapt slowest, 254 is adapt fastest, 255 is on and vendor specific. John Lohmeyer agreed with the "slider control" concept but objected to making 255 a special case. Doug Pickford questioned why we need more cache controls when existing customers appear not to be interested in supporting the existing controls. Gerry questioned whether the SLOG bit should be the on/off control for adaptive caching. The consensus was that SLOG should only control logging. Gerry's comments concerning the need for controls on the MODIFY DATA POINTER message were previously address at the Providence working group meeting. Gene Milligan has an action item to propose a common method of indicating which messages are supported. Gerry questioned whether we need controls for any other messages. There was agreement that we need two bits to control the MDP message, one for the read case and another for the write case, but agreement was not reached on which page should contain these bits. 2. Diagnostic Command Set (90-022) [Pickford] {Wednesday a.m.} Doug Pickford presented his revised proposal for a diagnostic command set. The new proposal is greatly simplified and fits better within the current SCSI command set architecture. Each item was separated into a separate proposal, so that the working group could discuss and accept/reject each without affecting the remaining proposals. John Lohmeyer requested a model be put in to establish the framework of the diagnostic functions. Dal Allan responded that the model is "conventional disk drive, period" and if someone reading this does not understand what that it is they should not be using the stuff. Later, after reviewing the proposals, the group concluded that a model would be useful to aid comprehension of several concepts including when the tools render the format corrupted. Proposal #1: Larry Lamers requested a read-only type designation for the cylinder descriptors. The seek bit caused some consternation among the group trying to understand what it meant. A drive with a dedicated servo may be able to seek to areas where the data area is not formatted for any use. After some discussion of the meaning of the seek bit, Doug agreed to fold the R/W and SEEK bits into a larger type field. Bob requested that safety and calibration areas be added to the type list. He also requested that two's complement notation be allowed for cylinder numbers so that negative cylinders can be described. He suggested bit seven of the new type field be a vendor specific indicator (that is, values greater than 127 be vendor specific). The starting and ending cylinder fields were extended to four bytes each. Starting and ending track fields were added to give track resolution for the areas. George Penokie questioned whether or not sector resolution is needed for identifying defect management areas. Proposal #2. The DS and TSD bits were removed from the log pages as defined in SCSI-2 to rationalize and simplify the logging operations. The group suggested that this be a separate proposal from the diagnostic proposal. Dal Allan stated that most of the information in this page would be far more valuable if the time duration were known. An elapsed time field will be added in units of minutes. George Penokie suggested the event count fields be expanded to four bytes. Proposal #3: This page defines the frequency of saving log pages, this eliminates the need for the DS and TSD bits. The working group recommended that this be included in a separate proposal. Proposal #4: John Lohmeyer stated the defined protocol of SEND/RECEIVE DIAGNOSTICS was not followed in the proposal. Doug agreed to change the proposal and add a corresponding RECEIVE DIAGNOSTIC RESULTS page. The starting logical block address and length were changed to starting cylinder, starting head, and number of tracks because this erase operation removes the logical block address. There was a discussion on how to get information about the erase operation. A value to be returned indicating a level of detected noise was proposed and will be included in the revised proposal. Proposal #5: Gerry Houlder asked how zones are handled. The response was that they are handled via the notched page mechanism in SCSI-2. It will be mandatory to place the sector components in sequential order. Pre-index, post-index or pre-id, post-id or pre-data, post-data gap will added to the list of component types. These are repeatable fields associated with a sector. John Lohmeyer asked why there is not a track composition descriptor. Doug plans to include revise the proposal to include track information. This may permit servo components to also be described. Proposal #6: The thresholds will be put into a separate page. Proposal #7: The purpose is to change the ID fields and not modify the gaps on a track. The layout of the page needs to be rationalized to the way diagnostic pages are organized. Gerry Houlder requested that the ID table entry fields be variable length with length fields. There was a question of the handling of split data fields and of redundant ID information in the headers. Sector components will be returned via the read track interleave operation. Chuck Micalizzi asked that a flag be added to the Format Log Page that indicates a format track operation has occurred. Proposal #8: It was suggested to change the name to "Write Sector ID" to more correctly describe the operation. Several people questioned the need for this operation and it was decided to drop this proposal. Proposal #9: Bob Snively asked how to handle drives that do not have an index. Every drive has some form of marker to indicate the start of the track. Doug will revise his proposal to not imply an index must exist. Overall Discussion: Doug Pickford asked if the group was still interested in a usage statistics page, for things like start/stop cycles. This page will be included in a separate log page proposal. Doug plans to revise his proposal and it will be on the agenda for the September working group meeting. Bill Spence thanked the group for its cooperation and its constructive input. 3. Proposed requirement for Diagnostic Command Set (90-89) [Snively] Bob Snively stated that Doug's latest proposal as modified is consistent with his recommendations. A key premise is that the drive is to be tested by trusted programs. Protection from less trustworthy programs should be done through vendor- specific protection mechanisms. Dal stated that the DCS stuff may destroy the ability to recover and service normal request without a low-level format. He suggested that we separate all commands into operational and diagnostic command sets. 4. Packetized SCSI (89-130R1) [Stephens] John Lohmeyer said that he is receiving management support in developing a draft SCSI FC-4 document that might be ready to submit sometime later this year. 5. Multi-ported SCSI (89-133R1) [Stephens] Deferred -- Gary Stephens was not able to attend. 6. 16/32-bit cable issues (90-48R1) [Penokie] George asked what the editor's intentions were on 32-bit options in SCSI-3. It seems that at present there is no clearly superior way to do 32-bit data transfers, therefore the editor was in favor of not including a 32-bit option in the initial draft of SCSI-3. 7. Can TDRD be used to block data transfer on current connection? (90-062R3) [Lohmeyer] John Lohmeyer presented his latest revision of the proposal which incorporates the results of the Wichita plenary meeting and proposed permitting the TTD message (was TDRD) to be used to block data transfers on subsequent connections. Dal Allan was opposed to adding this capability. He preferred to know on the first connection whether the target or the initiator would be responsible reconnecting for data transfers. John Lohmeyer pointed out that letting TTD messages block later data transfers would solve the overhead issue that Gerry Houlder had identified in his 90-83 document. Simplicity won out and John will prepare a rev 4 that only permits the TTD message on the initial connection. 8. SCSI-2 Request for Interpretation #1 (90-98) [Earle] {Should there be a Page Size Descriptor for each partition on the tape?} The group discussed the request and considered letters from Bharat Shah (90- 113) and Gary Stephens (90-109) on the topic. The conclusion was that the working group agreed with Brian Earle's opinion. The document seems clear that there are to be Page Size Descriptors for each partition, not just the additional ones. Since this was the first Request for Interpretation the group had ever received, John read the pertinent paragraphs from the X3 rules (SD-2). John suggested that the committee position should be that this be a class A response in that the standard is clear. John also pointed out that we are in an unique situation in that the rules assume the standard being interpreted has been published, but SCSI-2 has not been published yet. This may give us an opportunity to make editorial clarifications to portions of the standard that have caused confusion. Bharat Shah, Ed Young, and John Lohmeyer accepted an action item to draft a response. 9. SCSI-2 Request for Interpretation #2 (90-97) [Earle] {Should filemarks/setmarks be assigned block addresses?} Dal Allan cited Gary Stephens' document (90-109), paragraph 4 of section 9.1.6, and prior explanations of DDS to support the opinion that filemarks and setmarks are to be numbered. In particular, the last sentence from paragraph 4 of 9.1.6 says, "Filemarks and setmarks may or may not have recorded identifiers, but if identifiers are used in the format, then each mark is assigned a value even if it is not explicitly recorded." This seems to be a clear statement that filemarks and setmarks must be assigned block identifiers (i.e., block addresses). John pointed out that the description in the LOCATE command description is not as clear as in the model. He suggested that a reference to the model or some other clarification be placed in the LOCATE command description. Since this Request for Interpretation is also clearly covered by existing wording in the standard, John recommended that the committee position should be that this is also a class A response. Bharat Shah, Ed Young, and John Lohmeyer accepted an action item to draft a response. 10. Reduced Common-Mode Range on Differential SCSI (90-092) [Jessen] Erik Jessen presented a foil that illustrated two alternatives (New-1 and New-2) to reduce differential power requirements by reducing the common-mode voltage range. One of the alternatives (New-2) also reduced the Vod requirements from 1.0 volt to 0.5 volt. Now New-1 New-2 Units --------- --------- --------- ----- Voh (max) +12 +7 +7 Volts (min) +3 +3 +2.7 Volts Vol (max) +2 +2 +2.2 Volts (min) -7 -2 -2 Volts Vod +/- 1 +/- 1 +/- 0.5 Volts Terminator 150/330 150/330 122/550 Ohms Iterm 10/6 10/6 5/4 mA Zterm 122 122 100 Ohms worst case DC power 2.5 (5.0) 2.0 (2.3) 1.4 (1.7) Watts Infinite data xfer 1.7 (4.0) 1.4 (1.4) 0.9 (1.2) Watts 20% data xfer 1.1 (3.4) 0.9 (1.2) 0.6 (0.9) Watts Note: The power numbers only include the drivers and receivers. Those shown in () include protocol logic and high-impedance drivers. The worst-case power consumption could drop from around 2.5 watts to 2.0 watts by reducing the common-mode range and if the Vod is reduced, the power drops to about 1.4 watts. The reduced power consumption would permit placing differential drivers and receivers into a protocol chip with about 100-pins. The New-1 proposal is compatible with existing differential transceivers, but does not meet the existing standard. The New-2 proposal would work with existing differential devices, but would require that new terminators be installed using 122/550 ohm resistors. John asked if the outputs are shorted, do the drivers get hot and have a degraded life as is typical with other non-RS485 drivers. John pointed out that with SCSI, it is possible to short all the drivers at the same time if the cables are plugged in backwards. Erik responded that he is considering using a thermal shut-down circuit. Dal requested that the power dissipation be calculated using 0.5 volt Vod with the current terminators (New-3). The New-3 advantage would be that it could work with either terminator at lower power. The new terminator should give better signal quality because the impedance match is better. Active and passive zeros need further consideration. Erik had calculated his resistor values assuming a 5.0 volt TERMPWR. This needs to be re-done with 4.0 volts. The group also requested Erik to do these calculations assuming leakage current from 16 devices as would be the case when using a P cable. Erik plans to study the New-3 alternative further using the inputs from the working group. He expects to host an ad-hoc meeting on this topic early in August. Anyone interested in attending was asked to call Erik at his new phone number, 714-932-6612. 11. SCSI-2 Alt-2 Termination for the B Cable [Wicklund] Tom Wicklund was not present, but John Lohmeyer described Tom's concern. The Alt-2 termination figure only applies to the 50-pin A cable, and not the 68-pin B cable; it only shows 18 signals and the signal names are for the A cable. Tom's preliminary investigation of changes required for the circuit to work for a B Cable would include increasing the voltage regulator Imax from 600 mA to 1000 mA. Kurt Chan stated that the part is rated at 1.5 Amps with a heat sink. The decoupling capacitors may also need to be increased. It was agreed to clarify the figure in the document to indicate that it is only an example of an A cable alternative-2 termination. Dal Allan requested that the committee ask the terminator manufacturers to investigate alternative-2 P cable solutions. 12. SCSI-3 Physical Project Proposal (90-105) [Lohmeyer] John presented the proposal, the document will be in the mailing, and it is likely that a vote will be taken at the next plenary. There was sentiment expressed to relax the downward compatibility requirements. It would be a mistake to require complete downward compatibility. John agreed to delete the word "downward" from section 2.2. 13. Defect list length exceeding 64kb (90-110) [Penokie] The current media has about 1-2% defective blocks. With 2GByte drives imminent, the current READ DEFECT command cannot handle enough defects. It was suggested that the 12-byte READ DEFECT DATA(12) command from the optical memory device section accommodates 2**32 byte defect lists. John Lohmeyer pointed out that the FORMAT UNIT command only supports 2**16 byte defect lists. It will not be possible to issue a READ DEFECT DATA(12) command followed by a FORMAT UNIT command with a complete list of defect descriptors. Most people felt this was not a problem in that users should only add to the G list and should not use the complete list option. Robert Snively requested that the defects be paged (as is done for the READ BUFFER command data). It is not reasonable to have huge buffers to read the defect data into. A quick check indicated that this is feasible. George planned to update his proposal accordingly. 14. Request for clarification during logging operations (90-111) [Penokie] George presented a programming style analysis of reporting log exception conditions. He stated that he planned submit this document as an official request for interpretation. George led the group through the logic of his analysis and the group offered several suggestions for improving its clarity. George planned to revise the document and submit it in time for the July mailing. 15. Editorial clarification in SCSI-2 [Penokie] George Penokie pointed out two editing errors in rev 10c: 1) Section 4.8, paragraph 1, sentence 2. The phrase "...or equal to.." was omitted. 2) Section 5.6.21, Implementors note 2, sentence 4. The last phrase, "..or does..." conflicts with paragraph 2, sentence 2, of the same section. All of the phrase beginning with "or does" will be deleted. The English in the first part of this sentence will also be improved. The resulting sentence will be: "The correct method is to respond to an SDTR message with a MESSAGE REJECT message if the device does not support synchronous transfers." 16. High-density connector specification issues (90-107) [Chan] The retention clips on the high-density connectors are permitting these connectors to sag when mounted horizontally, especially with long cables. Ed Silva (Hewlett Packard) had pointed out this issue in the High-Density Interconnection Working Group Meeting at Wichita. This does not cause electrical discontinuities, but may deform the contacts which could result in damage on subsequent matings. The clips are not a problem if the connector is mounted vertically or in typical desktop systems where there is a small cable drop. Kurt Chan introduced 90-107 which describes the problem and speculates on possible solutions. Kurt stated that he had mixed feelings over how the committee should proceed. If we "clarify" the document to say the retention mechanism is optional, then we create a free-for-all in the marketplace. If we attempt to change to the jackscrew retention mechanism, then we definitely have a technical change on our hands. Kurt pointed out that the dimensions for the retention clips are not shown and that the three current vendors have slightly different clips. The clips inter-operate, but this aggravates the sag problem. The current jack screws and retention clips are mutually exclusive although at one time this was thought not to be so. Having a device connector that accepted either would be desirable. IPI and HiPPI have selected 2-56 jack screws made of high-tensile material. This maintains a good mechanical connection, but such a cable could pull a desktop computer off a table. The clips were designed to release before this would happen without damage to the clips. After some discussion the working group consensus was that the document should not be changed. 17. 1991 Meeting planning [Lohmeyer] John presented the proposed scheduling of the working group meetings for 1991. He had selected the dates to avoid having working group meetings so close after the plenary meetings as had been the case in 1990. He pointed out that this would mean the mailing deadline would occur prior to some working group meetings. Some conflicts with other industry events were noted. Alternative dates were selected. Dal and John planned to coordinate the dates with the HiPPI and Fiber Channel working groups. The dates are expected to be finalized by the Boulder plenary meeting. Anyone interested in hosting one of the working group meeting weeks in 1991 should contact John Lohmeyer at 316-636- 8703.