Date: January 15, 1990 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: January 9-10, 1990 X3T9.2 Working Group Meeting John Lohmeyer opened the meeting by thanking Bob Snively of Sun Microsystems for hosting the meeting in San Jose, CA. John noted that the Red Lion Inn had excellent meeting facilities, including generous A.C. outlets. The final agenda was as follows: 1. Develop alternative 16-bit and 32-bit solutions. (89-94R6, 89-150R1, 90- 004, 90-005, 90-006) [Penokie] 2. Draft responses on the public review comments on the 16-bit single- connector solution (P cable). {#3 & #4} (89-140 & 89-144) 3. Draft a resolve on the public review comment requesting a segment number field in the cache page. {#5} (89-152) 4. Draft a response to the public review comment requesting changes to the scanner command set. {#2} (89-158) 5. Draft a response to the public review comment requesting that a ribbon- style connector be added. {#6} (90-008) 6. Differential Transceiver discussion. [Murdock] 7. Review documents 89-133 and 89-130. [Stephens] 8. Reflections on Writing to EOT (X3T9.2/89-147 & 90-007) [Spence] 9. SCSI-1 and CCS Implementation on a SCSI-2 Target (90-003) [Houlder] 10. Clarification of Multi-Initiator Tagged Queuing (90-002) [Houlder] 11. Diagnostic Command Set [Dominguez] 12. Cable Specifications 13. SCSI-2 Editor Payment - SCSI-3 Editor Recruitment 14. Additional SCSI Caching Control (90-021R0) [Milligan] 15 General Working Group Schedule for 1990 The following people attended the meeting: Name Stat Organization ------------------------------ ---- ------------------------------ Mr. Denis D. Springer O 3M Company Mr. Al Wilhelm P Adaptec, Inc. Mr. Patrick E. Pejack A Adaptec, Inc. Mr. Danial Faizullabhoy O Adaptec, Inc. Mr. Kheng Bin Ng O Adaptec, Inc. Mr. Bob Whiteman A AMP, Inc. Mr. Peter M. Blackford P Astro Cable Company Mr. Joe Lawlor P AT&T Mr. John A. Gibson P Berk-Tek, Inc. Mr. Wills Xu O C&M Corp. Mr. Mark P. Pearson P Datacopy Corp. Mr. John D. Walden P E.I. DuPont Inc. Mr. Stephen L. Clark A E.I. DuPont Inc. Ms. Jean Kodama S Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. John C. Onia V Fujitsu America, Inc. Mr. Terry Maezawa O Furukawa Electric Amer, 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. Gary R. Stephens A IBM Corp. Mr. Hubert Lee V IBM Corp. Mr. Bill D'Andrea V JST Corporation Mr. Erik Walberg O Konica Technology Corp. Mr. William Homans P LMS-TSD Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Richard Wagner O Montrose Products Company Mr. James Schuessler P National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. David Steele S NCR Corp. Mr. Dennis P. Trupski P Olivetti Mr. Wayne Baldwin O Olivetti Mr. Gerald Houlder O Seagate Technology Mr. Gene Milligan O Seagate Technology Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. Dexter Anderson O Sun Microsystems, Inc. Mr. Vit Novak O Sun Microsystems, Inc. Mr. Matthew Tedone O Sun Microsystems, Inc. Mr. Rich Clewett V Sun Microsystems, Inc. Mr. Pete Tobias A Tandem Computer Inc. Mr. D. W. Spence P Texas Instruments Mr. Ricardo Dominguez S Texas Instruments Mr. Erik Jessen O Western Digital 45 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 ------------- -------- --------------- --------------------------------------- X3T9.2/89-94 12/28/89 G. Penokie 16 Bit Data Path on a Single 68-pin Rev 6 Connector X3T9.2/89-150 12/28/89 G. Penokie 16 Bit Data Path on a single 68-pin Rev 1 Connector (Appendix form) X3T9.2/89-152 12/1/89 Robertson/ SCSI-2 Public Review Comment #5 (re: Cornaby Caching) X3T9.2/89-158 10/23/89 P. Joslin SCSI-2 Public Review Comment #2 (Re: Scanner Cmnd Set) X3T9.2/90-2 12/22/89 G. Houlder Clarification of Multi-Initiator Tagged Queuing X3T9.2/89-3 12/22/89 G. Houlder SCSI-1 and CCS Implementation on a SCSI-2 Target X3T9.2/89-4 12/28/89 G. Penokie 32-bit Data Path on a Single 110-pin Connector (L Cable) X3T9.2/89-5 12/28/89 G. Penokie 16/32-bit P/Q Cable Stand Alone Document X3T9.2/89-6 12/28/89 G. Penokie 16/32-bit P/L Cable Stand Alone Document X3T9.2/89-7 1/9/90 B. Spence Proposal for a Note re Writing to EOP (PEOT) X3T9.2/89-8 11/30/89 R. Graczyk SCSI-2 Public Review Comment #6 (re: High Density Connector) X3T9.2/90-16 1/5/90 C. Grant Near End Crosstalk Measurement X3T9.2/90-17 12/20/89 J. Gibson Proposed Near End Crosstalk Measurement Procedures X3T9.2/90-18 J. Filia Cable Test Methods X3T9.2/90-19 J. Filia Impedance Nomograph X3T9.2/90-20 1/4/90 K. Chan Shielded Cable TDR Measurements X3T9.2/90-21 1/8/90 G. Milligan Additional SCSI Caching Control X3T9.2/90-22 12/11/89 D. Pickford SCSI Diagnostic Model - Rigid Disk Drives Results of meeting 1. Develop alternative 16-bit and 32-bit solutions. (89-94R6, 89-150R1, 90- 004, 90-005, 90-006) [Penokie] The 89-094R6 document changes the Q cable pin-out to make it pin compatible with the P cable thus allowing for the possibility of dual porting. It also disallows intermixing of A and B cabled systems with P and Q cabled systems. The 89-150R1 document is a version designed to be included as an appendix with same changes as 89-094R6. The 90-004R0 document is the same as 89- 150R1 except that it drops the Q cable and adds a 32-bit L cable. The 90- 005R0 document is a stand alone version for P and Q cables. The 90-006R0 document is a stand alone version for P and L cables. The pros and cons of single-cable (L) and dual-cable (P&Q) 32-bit wide SCSI was discussed. The negatives of the L cable are: 1) L cable is much stiffer, requiring a larger bend radius, but only 0.1 inch larger in diameter than the Q cable. 2) L cable is heavier and may require a better retentions systems, such as jack screws. 3) L cable may require bulkhead mounting. 4) L cable connector is too large for many expansion slots. 5) L cable may have poor electrical characteristics, but no testing has been done. The negatives of the Q cable are: 1) Q cable requires two cables and two connectors. 2) Q cable may have crosstalk problems. 3) Q cable needs to have dual REQ/ACK and the logic to re-assemble skewed data from the two cables (same problem as in A/B). 2. Draft responses on the public review comments on the 16-bit single- connector solution (P cable). {#3 & #4} (89-140 & 89-144) John Lohmeyer presented draft responses to these public review comments (see attachments). The proposed responses reject the idea of delaying SCSI-2 publication in order to include a 16-bit single-cable solution. The working group recommends that X3T9.2 adopt the proposed responses by a straw poll with 9 in favor, 3 opposed, and 2 abstentions. 3. Draft a resolve on the public review comment requesting a segment number field in the cache page. {#5} (89-152) John Lohmeyer presented a draft response to this public review comment (see attachments), which would delay the proposed change until SCSI-3. This would also permit the committee to investigate other solutions. The working group recommends that X3T9.2 adopt the proposed response. 4. Draft a response to the public review comment requesting changes to the scanner command set. {#2} (89-158) John Lohmeyer presented a draft response to this public review comment (see attachments), which would delay the proposed changes to the scanner command set until SCSI-3. The working group recommends that X3T9.2 adopt the proposed response. 5. Draft a response to the public review comment requesting that a ribbon- style connector be added. {#6} (90-008) John Lohmeyer presented a draft response to this public review comment (see attachments), which rejects the comment. The working group recommends that X3T9.2 adopt the proposed response. 6. Differential Transceiver discussion. [Murdock] Gary Murdock was not in attendance. He sent word that there is no additional information available at this time. He plans to attend the next meeting and bring the information. 7. Review documents 89-133 and 89-130. [Stephens] Gary's presentation engendered some heated discussion about terminology, in particular the exact meaning of hosts, initiators, and targets. Trying to discuss dual-porting with the existing terms is nearly impossible since those terms are not precisely defined enough in SCSI-2 to deal with the situation. One needs to get on a conceptually higher plane to grasp the possibilities and ramifications of a dual-ported SCSI. In reading Gary's proposals, it may be helpful to note that Gary has defined the terms initiator and target in a fairly narrow sense. These are the portions of the device associated directly with communication with an individual SCSI bus. Gary's initiator increments pointers as data is transferred, but the pointers reside above the initiator as part of the host adapter. This permits another initiator on the same device to continue an I/O process if something makes the original initiator unavailable. Gary's target can be thought of as a port to the device controller. There may be more than one target on each device controller. The device controller could receive an I/O process on one target port and continue the I/O process on another target port. Dal, in his usual polite fashion, asked Gary to pick some new terminology for subsequent revisions to his proposals. 8. Reflections on Writing to EOT (X3T9.2/89-147 & 90-007) [Spence] Bill Spence has been writing proposals on this topic for some time. He finally found wording that was acceptable to the group. Bill's note was accepted as modified for incorporation in the SCSI-2 document as a non- substantive change. 9. SCSI-1 and CCS Implementation on a SCSI-2 Target (90-003) [Houlder] Gerry Houlder has attempted to define explicitly the affect a CHANGE DEFINITION command would have on a target device. This topic generated an interesting and lengthy discussion on how well SCSI-1 initiators would survive in an SCSI-2 target world. The list of real problems seems to be relatively short: 1) SCSI-1 targets respond to single ID selections while SCSI-2 targets do not (be may do so to emulate SCSI-1) 2) INQUIRY data returned by an SCSI-2 target may confuse an SCSI-1 initiator 3) a REQUEST SENSE command with the allocation length set to zero may cause problems because the SCSI-1 initiator would expect four bytes of sense data 4) tape devices under SCSI-2 may behave somewhat differently (more data is needed for this item). Gerry plans to revise his proposal and return for another round. 10. Clarification of Multi-Initiator Tagged Queuing (90-002) [Houlder] The current SCSI-2 document prohibits untagged and tagged queuing being used at the same time on a target. It doesn't say what should happen if one initiator is using tagged queuing and another initiator tries to use untagged queuing. Gerry Houlder's proposal would "clarify" the document to say that the prohibition against mixing untagged and tagged commands is on a "per initiator" basis. Larry Lamers brought up the paragraph about the Queue Algorithm Modifier in section 7.3.3.1. His opinion was that this paragraph implies that the command queue is on a target basis and that when the Queue Algorithm Modifier is zero the target must appear to execute commands in the order received, regardless of which initiator issued the command. This was a lengthy discussion between Gerry Houlder, Dal Allan, and Larry Lamers on the meaning of Queue Algorithm Modifier. As written, the Queue Algorithm Modifier, when set to zero, guarantees data integrity; that is, the target executes the tagged commands in such a manner that the observed changes to the medium would be the same as if queuing was not used. What is interesting about this is that there is no such requirement for untagged queuing. And, apparently, there was no intention to require the target to maintain strict time-ordering of commands from different initiators in a tagged queuing environment. Note that guaranteeing correct time-ordering of the I/O processes is not really a guarantee of file integrity -- systems must include record locking facilities to insure correct files. The resolution of the discussion was that the working group recommends X3T9.2 make the proposed change to section 6.8 except for the last sentence. In addition, Gerry Houlder, John Lohmeyer, Dal Allan and Larry Lamers accepted an action item to draft a proposal to clarify the architecture of command queuing to a per-initiator basis as it relates to the Queue Algorithm Modifier. 11. Diagnostic Command Set [Dominguez] Ricardo Dominguez was called away before the discussion could commence. The latest document, authored by Doug Pickford, (see 90-022R0) will be distributed in the mailing along with a cover letter from Ricardo. Robert Snively "calmly" suggested that the document be rejected because section 2.1.5, which would permit both DATA OUT and DATA IN phases during the same I/O Process, is a conceptual violation of the SCSI protocol. 12. Cable Specifications Larry Lamers reported on the Ad-hoc Cable meeting held on Monday January 8, 1990 (see 90-023). The recommendations from the cable working group are: 1) that the 90 ohm minimum impedance requirement be relaxed to a recommendation and that a warning be included 2) that a warning on attenuation, rise time, and noise parameters affecting cable performance be included in SCSI-2 3) that it be made clear that the 28 AWG minimum wire size requirement applies only to TERMPWR and that the 90 ohm impedance applies only to signal pairs. This allows cables to be built that use 30 AWG wire for signals which improves the characteristic impedance. Several members of the group met later in the day to draft notes for inclusion in the document. Bill Spence presented the proposed notes. Dick Wagner of Montrose Cable presented a re-write of paragraph 4.2. There was agreement that paragraph 4.2 needed some wording changes. John Lohmeyer presented an additional implementors note. This note was accepted for inclusion in SCSI-2. Gary Stephens asked whether there were two vendors of cables that could meet the SCSI-2 specification for 90 ohms. The cable vendors responded that the 90 ohm specification could be met on data signals using 30 AWG wire, however the complexity of manufacturing mixed wire diameter cables would result in higher cost cables. John Lohmeyer, Bill Spence, George Penokie, and Larry Lamers drafted revisions for section 4.2 that will be recommended to X3T9.2 for incorporation in SCSI-2. These changes are considered non-substantive in that no new requirements are added and one requirement (the 90 ohm minimum impedance) is relaxed to a recommendation. 13. SCSI-2 Editor Payment - SCSI-3 editor recruitment X3T9 has received tentative approval from CBEMA to set up an account that could be used for such things as paying an editor on the SCSI-2 document. Collecting funds cannot be made a condition for membership in X3T9.2. Contributions must be voluntary. The question was raised, "What happens if we do not convert the document and let ANSI do the work?" John Lohmeyer replied that minimally the cost of the standard would increase to cover ANSI costs and we would lose control over the technical content. John accepted an action item to investigate the prospect of ANSI paying for part of the editors cost since they are responsible for publishing standards. NOTE: Del Shoemaker, X3T9 chairman, went into hysterics when asked this question. It appears to be out of the question -- John Dal Allan accepted an action item to investigate the possibility of having Keith Brannon's editors at ISO do the conversion. 14. Additional SCSI Caching Control (90-021R0) [Milligan] Gene Milligan presented a SCSI-3 proposal for extending the cache control page to support segments and to enhance the feature control of caches. The working group suggested that clarification is needed regarding the relationship between the Cache Segment Size field and the FIFO Segment Size field. Gene plans to revise the document. 15 General Working Group Schedule for 1990 The 1990 Working Group Schedule is as follows: Date Location Host Comments ------------------- -------------- --------------- ---------------- Mar 5-9, 1990 Costa Mesa, CA Western Digital Red Lion Inn May 7-11, 1990 Providence, RI Astro Cable Marriott Jul 9-13, 1990 Rochester, MN IBM {tentative} Sep 4-7, 1990 Denver, CO Storagetek {no CAM meeting} {tentative} Oct 29-Nov 2, 1990 Austin, TX TI {tentative} These week-long meetings are to be divided as follows: Monday SCSI CAM Committee (except September 1990) Tuesday-Wednesday SCSI Working Group Thursday-Friday Fiber Channel Attachments: Draft Responses to the SCSI-2 Public Review Comments The following draft letters were reviewed by the working group and are recommended to X3T9.2 for resolution of the SCSI-2 comments. Only the contents of the letters are included here: Public Review Comment #2 Dear Mr. Joslin, Thank you for your interest in the SCSI-2 draft proposed standard. The X3T9.2 committee has reviewed your public review comment (X3T9.2/89-158) suggesting several changes to the Scanner Command Set to accommodate your Photomatrix Aperture Card Scanner product. Many of your ideas have merit, however the Aperture Card Scanner does not fit the Scanner Device model in SCSI-2, which was principally developed for page scanners. As such, X3T9.2 does not feel it is appropriate to incorporate the changes in SCSI-2. The X3T9.2 committee has recently begun working on SCSI-3. We would like to encourage you to participate in this project. It may be possible to enhance the Scanner Device model in SCSI-3 to include such devices as your Aperture Card Scanner. Also, some of the changes you suggested may be useful for next generation page scanners. If you need more information concerning participating in X3T9.2, please contact John Lohmeyer, X3T9.2 chairman, at 316-636-8703. In the meantime, X3T9.2 recommends that you use the SCSI-2 command set with appropriate vendor-specific extensions where necessary. Public Review Comment #3 Dear Mr. Stai, Thank you for your public comment (X3T9.2/89-140) concerning the SCSI-2 draft proposed standard. While the X3T9.2 committee generally agrees with your position that in the future the SCSI-3 16-bit "P" cable solution will likely be more widely implemented than the SCSI-2 16-bit "A/B" cable solution, the committee feels that the "P" cable is not sufficiently stable at this time to include it in SCSI-2. In particular, the method by which a "P" cable system may be expanded to 32 bits needs more study. Furthermore, it would be inappropriate to remove documentation of the "B" cable when some products have already implemented the "B" cable. X3T9.2 plans to have a single-connector 16-bit solution in SCSI-3 including a method of expanding to 32 bits. A draft working document is expected later this year. In order to alert SCSI-2 implementors of the potential for changes in this area, a note will be included in SCSI-2 saying that X3T9.2 is documenting an alternative 16-bit single-cable solution and an alternative 32-bit solution and expects to be able to remove the "B" cable in a subsequent version of SCSI. Public Review Comment #4 Dear Messrs. Robertson and Cornaby, Thank you for your public comment (X3T9.2/89-144) concerning the SCSI-2 draft proposed standard. The 16-bit single-cable option is not included in SCSI-2 because the X3T9.2 committee feels that it is not sufficiently stable at this time. In particular, the method by which a 16-bit single-cable system may be expanded to 32 bits needs more study. X3T9.2 plans to have a single-connector 16-bit solution in SCSI-3 including a method of expanding to 32 bits. A draft working document is expected later this year. Your suggestion that we avoid major modifications to the body of the SCSI-2 standard by documenting the "P" cable in an appendix is appreciated. However this would not avoid the need for another public review period and its attendant delay in publication of SCSI-2. X3 considers the addition of such a major feature as a "substantive" change, no matter whether it is done in the body of the standard or in an appendix. In order to alert SCSI-2 implementors of the potential for changes in this area, a note will be included in SCSI-2 saying that X3T9.2 is documenting an alternative 16-bit single-cable solution and an alternative 32-bit solution and expects to be able to remove the "B" cable in a subsequent version of SCSI. Public Review Comment #5 Dear Messrs. Robertson and Cornaby, Thank you for your public comment (X3T9.2/89-152) concerning the SCSI-2 draft proposed standard. The X3T9.2 task group has reviewed your comment and intends to address your concern as part of the SCSI-3 project, which has already started. A draft SCSI-3 working document is expected later this year. Public Review Comment #6 Dear Mr. Graczyk, Thank you for your interest in the SCSI-2 draft proposed standard. The X3T9.2 committee has reviewed your public review comment (X3T9.2/90-008) suggesting that a high-density ribbon contact connector be included in the standard. The committee has devoted considerable attention to the connector issue. You mention usage problems with some pin-and-receptacle connectors as the reason the "large" ribbon connectors were developed. As you probably know, SCSI-1 included a miniature ribbon connector (0.085-inch contact centers) after problems with its original pin-and-socket connector became apparent. The X3T9.2 committee certainly shared your concern when selecting the high- density SCSI-2 connector. The fact that they selected a tab-and-receptacle contact design in spite of the previous experience serves to demonstrate the robust design of these high-density connectors. They have been in production for several years and are available from at least three vendors. No usage problems have been reported to the committee. Connector selection for I/O interface standards is always a difficult and time-consuming process. X3T9.2 had many excellent connector proposals from which to select, including several that employed ribbon contacts. After a successive elimination voting procedure, the current high-density connector was selected. The final design was endorsed by a roll-call vote which passed 48 Yes, 7 No, 0 Abstain, and 13 Absent. The committee has also consistently rejected the idea of including more than one high-density connector design in the SCSI-2 draft standard. This would weaken the standard and create market confusion. Accordingly, X3T9.2 has voted against adding a ribbon-contact connector to the SCSI-2 standard.