Ms. Lynn Barra November 3, 1989 CBEMA X3 Secretariat 311 First Street NW, Suite 500 Washington, DC 20001-2178 Subject: Public Comment on BSR X3.131-198X, Small Computer Systems Interface (SCSI-2) It has come to our attention that the Task Group X3T9.2, which is already working on extensions to the SCSI-2 standard, has developed an alternative "wide" data path (known as the "P" cable) to those already defined for SCSI-2. This new "P" cable has been developed for inclusion in a future SCSI-3 standard. We feel that this new data path, which performs a function identical to a "wide" data path defined for SCSI-2 (known as the "B" cable), will cause confusion in the marketplace and therefore cause a delay in the acceptance of devices that will transfer data on a "wide" data path. Currently available SCSI designs that conform to X3.131-1986 transfer data on an 8-bit cable, known now as the "A" cable. X3T9.2, as part of its SCSI-2 effort, came up with a method to transfer data at faster rates using a wider data path. This wider data path is now known as the "B" cable. The "B" cable is used to transfer either 16 bits or 32 bits in conjunction with the "A" cable. For 16-bit transfers the "B" cable carries only 8 bits; the other 16 bits on the cable are not used. For either 16-bit or 32-bit operation, the device must have a 68-pin "B" cable connector, and a 50-pin "A" cable connector. At the time this system was developed by X3T9.2, this arrangement seemed appropriate since the devices that would carry these connectors would be large enough to contain them. Today, with device sizes shrinking rapidly, this is no longer the case. This is why the "P" cable, which can carry 16 bits on a single 68- pin connector, is now being developed. The marketplace confusion arises in the future when 16-bit "P" cable systems are to be upgraded to 32-bit operation with 32-bit "A/B" systems. The only viable alternative for those devices to transfer 32 bits is to add a "B" cable connector to the "P" cable device. The problem with this is that the "B" cable must carry 24 bits to be compatible with the "A/B" system, while the "P" cable carries only 8 bits. This means that the "P" device is wasting half its resources. Granted, this is manageable, but it is not desirable to a device manufacturer, since we are shipping circuits and pins that are never used, needlessly increasing the system cost. A more fundamental problem occurs in the above scenario when 16-bit "A/B" devices are connected. We now have a 32-bit "A/B" device, a 16-bit "A/B" device, and a 16-bit "P" device. The two "A/B" devices can transfer data successfully with each other, but when the "P" device attempts to transfer data with the 16-bit "A/B" device, it has no way in the current protocol to determine how to transfer 16-bit data. The danger here is that SCSI-2 "A/B" protocol would have to be modified to handle this case in SCSI-3, causing those SCSI-2 devices to become obsolete. It is apparent to us that today's design constraints favor the use of the "P" cable solution for wider data transfers. It is also apparent to us that the "A/B" approach is not desirable and may be harmful to the continued growth of the standard. Therefore, we ask that X3T9.2 consider removing the "B" cable from SCSI-2, and add the "P" cable solution in its place. We believe that the "pain" of delaying the approval of SCSI-2 will be well worth benefit of quicker market acceptance of a single solution for wide data transfers. Thank you for your consideration of our comments. _____________________________ Jeffrey Stai for Western Digital Corporation 2445 McCabe Way Irvine, CA 92714 CC: ANSI Board of Standards Review 1430 Broadway New York, NY 10018 Mr. John Lohmeyer, X3T9.2 Chairperson NCR Corporation 3718 North Rock Road Wichita, KS 67226 Mr. Del Shoemaker, X3T9 Chairperson Digital Equipment Corporation 1331 Pennsylvania Ave NW Washington, DC 20004