Date: 5/31/93 X3T9.2/93-103r0 To: R. Fink Cc: SCSI Reflector From: G.E. Milligan OKM251 Subject: SCSI-3 Parallel Interface Letter Ballot Comments Unfortunately I left my letter ballot in the pocket of an airplane seat. As I recall the ballot was closing about now. I vote YES with a number of comments. Reflecting upon the greater course of X3T9 I note that we have come through several major stages. Initially many participants were attempting to block standards since they would impede technology development. Customers made them see the light and the next stage was the renaissance of standards development. Participants sought out the optimum standards and worked diligently towards their completion. We have now progressed(?) into the gluttony of standards. I/O definitions which would otherwise be proprietary interfaces are taken out shopping to find a standards body which will add on another project to document yet another interface without regard to how it fits within the multiplicity of alternatives. All of this has attracted so much participation that the many projects seem to never end. A large portion of the participation is on an intermittent or selective location basis. Consequently the development process stretches out interminably. Since the formal process coupled with the uncontrolled participation doesn't fit with modern product life cycles changes in the process are required. Several of the committees have used the ploy of forwarding a dpANS prematurely to hold a sham public review to get the longer four months review out of the way while the real dpANS is completed. The track record on this method does not suggest that it should be supported. A better policy/procedure which I advocate is to recognize that successful standards are living documents which will require periodic updating. They should be processed expediently, published, and undergo frequent (12-18 month updates) updates as the technology evolves. The SCSI-3 SPI is one of these living documents. The technical issues have been thoroughly aired and decided to the ability of X3T9.2 and it should not be returned to X3T9.2 for further delay in being synchronized with the new generation of SCSI products. The wide version of SCSI in SCSI-2 is horribly obsolete and needs to be replaced by the definition in SCSI-3 SPI. Any new technical issues can be decided and published in a newer version. This does not mean that there should be no changes before publication. There have been a number of issues long since decided that need to be better reflected in the document. The editor has done a very good job of documentation but some items do not yet reflect the sense of the committee. Along these lines I offer the following specific comments: 1) On page i make the last line of the Abstract "... both system integrators and ...". 2) Delete page ii Document Status and move the points before "Revision 7" to the Introduction or Scope. Before moving it, change "several changes" to "several changes or additions" and change item 1 to "Elimination of B cables." The committee has repeatedly confirmed that A cables are still addressed by SCSI-3. Remove the period from item 3 or add periods to the other items. 3) In the table of contents and in Annex F Figures E1 and E2 should be F1 and F2. 4) Add "(SPI)" after "SCSI-3 Parallel Interface" in the Foreword on page ix. 5) In the first line of the Introduction on page x replace "with up to 32 devices" with "with up to 8, 16, or 32 devices depending upon the data path widths implemented". 6) On page 2 delete everything in the first paragraph after "This roadmap is intended to show the general applicability of documents to one another" including the deletion of the three bulleted items. 7) On page 4 in the second paragraph delete "(ISO 8482-1982 TIA TR30.2)" as it has nothing to do with that paragraph. 8) Add titles to IEEE 1156.2 and ASTM D-4566. Add a designation to the drafting standard. 9) On page 5 the differential definition implies there are two different differential definitions. Change "One of two signaling alternatives" to "A signaling alternative". Make the analogous change to 3.1.27. 10) The REQ/ACK handshakes are not necessarily synchronous. In 3.1.7 delete "that uses the synchronous REQ/ACK handshake" or delete the second synchronous. 11) In 3.1.8 delete "and" or replace it with "which". 12) Change "rate that bytes of data" to "rate that words of data" or "rate at which words of data". 13) Expand the optional definition 3.1.14 to include "but if the item is implemented in shall conform with the definitions in this standard." 14) Modify 3.1.21 from "(0-31)" to "(0-7, 0-15, or 0-31 depending upon the data path widths implemented". Make the analogous change to 3.1.22. 15) Modify the SCSI definition in 3.2 to include SCSI (or SCSI- 1). 16) It does not seem appropriate to define "PPB" and "MFG" since they are used only one place in the standard. (Due to all the work including the place they are used I will refrain from commenting that the inclusion is inappropriate in the SCSI standard.) 17) The second paragraph on page 8 appears to be out of place. Perhaps it should be in section 10. 18) Delete the second sentence in the third paragraph since it is anecdotal and not part of the standard. 19) In several places beginning with the second to last paragraph on page 8 replace "asynch" with "asynchronous". 20) In that same paragraph replace "has occurred" with "is in effect". 21) The last paragraph on page 8 is inaccurate. The annexes cover more than measurement techniques. A broader description is needed. 22) In 5.1 change the first sentence to "The P cable nonshielded SCSI-3 connector ...". Make the analogous change to 5.2. 23) The d.c. resistance on page 10 is in conflict with the passing criteria of section 5-5. Increase it to at least 15 milliohms and I presume it should be at least 25 milliohms. 24) All of the Host Bus Adapters I have seen violate the Note on page 10. Why have the HBA companies not objected? 25) Change the title of Figure 4 to "P cable Non-shielded device connector" or "P and Q ...". Make the analogous change to Figures 5, 6, and 7 along with Tables 1, 2, 3, and 4. The equivalent change needs to be made to the first sentence in sections 5.3, 5.3.1 (and the second sentence), and 5.3.2 (and the second sentence). 26) Change the fourth paragraph on page 19 to "A primary SCSI bus carries a 8 or 16-bit data bus ...". Change the last sentence of that paragraph to "The P cable/connector ... buses." Beyond distinguishing the P cable from the A cable this change also eliminates "using either single ended or differential transceivers" which give the mistaken impression that the P cable can be single ended while the Q cable is differential. 27) The note callout "N" in section 6 is incorrect and the grammar is wrong. I guess it is not referring to Annex A and I presume it should instead reference SCSI-2. 28) The paragraph between the two notes in section 6 is misleading. Since this section addresses cables the statement that they should be terminated at both ends is wrong. The termination description should be moved or made clear that the termination is at the ends of the total bus including all cables. 29) The second note is misleading. The device with the terminators should not be removed even if the device is not in use since the terminators are presumably still in use. 30) The editors note [ljl] should be deleted and perhaps brought to fruition. 31) In the first sentence of 6.1 delete the reference to 8.1. If the reference is desirable move it to section 6(.0). 32) Change the last item in Table 5 to "... signals in the same cable". 33) Note that the first line of 6.3 is in direct conflict with the note on page 10. 34) I presume in the second paragraph of 6.4 the requirement is stated twice. Delete "and stub clustering avoided". Also delete it in section 6.5. 35) In the last paragraph replace "on interconnection" with "on interconnecting buses of different widths" and "extended length operation" with "terminator, impedance, crosstalk, and bus length considerations". 36) It is a compete puzzle as to what the note at the bottom of page 21 is referring to. If I had to rival Mycroft I would presume it is a mistaken characterization of the A cable as SCSI- 2 and the P cable as SCSI-3. The best solution is to delete the note. 37) Does the requirement of 7.1.1 (b) allow the signal voltage to be at least 2.5 v as in (c) or does it just allow it to be 2.5 v max? Active negation would allow at least 2.5 v. 38) It has been alleged that the requirements of 7.1.2 are stated in the manner clearest to semiconductor engineers. Skilled engineers seem to stumble over the description. It should either be turned inside out or a sample load line diagram should be given so the real requirement is revealed to other than the working group participants. 39) In the second paragraph of 7.1.3 move "(e.g. with some ESD protection circuits)" to immediately after "occur". Also delete "does" from "does should not". 40) Delete the note on page 26 since its only purpose seems to be to say that the standard should not be violated. Without the note it seems mandatory to not violate the standard. 41) Table 7 is not clear as to what the requirement is for the SCSI-3 A cable implementations. The note below may have started to give the requirement but unfortunately the author died before finishing the sentence. 42) In section 8 change the first sentence to "... maximum of 8 SCSI devices on an A cable, 16 ...". 43) In the note at the bottom of page 28 change "the margin is much higher" to "the margin is higher". 44) In Table 9 in the "None:" note delete "be" from "not be driven by any SCSI device." 45) Sections 9.8 and 9.10 specify specifically how to measure timing for single ended. Shouldn't an analogous requirement be included for differential? 46) The requirement of 10.2.1 is ok. However I note that it would become in conflict with the "Son of Spastic" proposal now being discussed. 47) In 10.3.1 change item (3) from "the SCSI ID" to "its SCSI ID". 48) At the top of page 40 and in 10.8.4 change "When target PIA" to "When the target PIA". 49) In 10.11 the acronym RAT has no convincing correlation with what it stands for. 50) Should the first MESSAGE in Table 12 be MESSAGE OUT? 51) The note on page 47 was appropriate in SCSI-2. Should it be retired in SCSI-3? 52) The two "shalls" should be deleted from Annex A. 53) Change the impedance in Annex D from "73 ohms" to "72 ohms" to match Table 5. 54) Under the heading of beating my head against a wall, I again take the opportunity to point out the "an SCSI" is an abortion and should be replaced with "a SCSI". Why cater to those who do not know how to pronounce "SCSI"? G.E. Milligan