(Replying directly to increase the chances you see this.) I received a couple more responses, from Sunil Sarin and from Mark Dewitt. Sunil had this to say, which was certainly surprising to me:
> John, I'll have to look in my hard copy archives, which are very disorganized and will take me some time. But I don't believe you need this exact paper. I can probably point you at other papers that were in fact published (and are not just company TRs).
> Furthermore, our concoction of the acronym SHARD is different from "database sharding" as currently used. We were referring to the network, not the data, being "sharded", i.e., partitioned, and ensuring high database availability (with some loss of consistency - see "The CAP Theorem").
Mark Dewitt shared:
> Hi John, I believe that was an unpublished report we wrote for our DARPA funding agency. I may still have a copy, but it's currently buried under some stuff in the garage. There is at least one published paper that describes the SHARD architecture for a partially replicated database. It's not entirely obvious because I think the SHARD acronym didn't start getting used until after the paper was published in 1985. By 1987, Sunil Sarin had published another paper with Nancy Lynch in which they refer to SHARD and reference the 1985 paper.
> Here are the two citations:
> S. K. Sarin and N. A. Lynch, "Discarding Obsolete Information in a Replicated Database System," in IEEE Transactions on Software Engineering, vol. SE-13, no. 1, pp. 39-47, Jan. 1987, doi: 10.1109/TSE.1987.232564.
> S. K. Sarin, B. T. Blaustein and C. W. Kaufman, "System architecture for partition-tolerant distributed databases," in IEEE Transactions on Computers, vol. C-34, no. 12, pp. 1158-1163, Dec. 1985, doi: 10.1109/TC.1985.6312213.
> John, I'll have to look in my hard copy archives, which are very disorganized and will take me some time. But I don't believe you need this exact paper. I can probably point you at other papers that were in fact published (and are not just company TRs).
> Furthermore, our concoction of the acronym SHARD is different from "database sharding" as currently used. We were referring to the network, not the data, being "sharded", i.e., partitioned, and ensuring high database availability (with some loss of consistency - see "The CAP Theorem").
Mark Dewitt shared:
> Hi John, I believe that was an unpublished report we wrote for our DARPA funding agency. I may still have a copy, but it's currently buried under some stuff in the garage. There is at least one published paper that describes the SHARD architecture for a partially replicated database. It's not entirely obvious because I think the SHARD acronym didn't start getting used until after the paper was published in 1985. By 1987, Sunil Sarin had published another paper with Nancy Lynch in which they refer to SHARD and reference the 1985 paper.
> Here are the two citations:
> S. K. Sarin and N. A. Lynch, "Discarding Obsolete Information in a Replicated Database System," in IEEE Transactions on Software Engineering, vol. SE-13, no. 1, pp. 39-47, Jan. 1987, doi: 10.1109/TSE.1987.232564.
> S. K. Sarin, B. T. Blaustein and C. W. Kaufman, "System architecture for partition-tolerant distributed databases," in IEEE Transactions on Computers, vol. C-34, no. 12, pp. 1158-1163, Dec. 1985, doi: 10.1109/TC.1985.6312213.