Summary

  • RFC 1357 turned the mailed announcement of a technical report into a bibliographic email that a catalog could parse. Its sender identified the report; the receiver chose whether to add the record.
  • Corrections and withdrawals arrived as complete replacement records with a later revision number. The design spared libraries from reconstructing a report's current description from a broken chain of messages.

Consider the work behind one library search result. An institution announces a technical report. A second institution receives the announcement, decides to admit it, and makes a card that its readers can find. Under postal distribution, the first step consumed days and the second required transcription. Danny Cohen's RFC 1357, published in July 1992, replaced the paper announcement with a plain-text email. It did not turn many catalogs into one database; it made the handoff between them legible.

The handoff soon became real infrastructure. The successor RFC 1807 states that its revised format was used by Cornell University's Dienst protocol and Stanford University's SIFT system. Those named systems give the format a job beyond the sample page: a technical-report service can send records that another service understands without agreeing to surrender control of its own index. RFC 1807 also describes work on converters to and from library cataloging records. The bridge between a research report and a library search had become worth standardizing twice.

RFC 1357's first four required fields reveal what the sender had to supply. BIB-VERSION identified the format. ID joined the publisher's controlled symbol to its report number with //. ENTRY dated the record, and END closed it. They appeared in that order. Title, author, abstract and retrieval instructions were optional. A library could use the stable publisher-and-report key to match a new message to a card it already held, while still applying its own admission rules.

The second message was where the design earned its keep. A corrected retrieval address or title could arrive after the original record had been copied into several catalogs. REVISION carried a higher integer and an explanation, but the revised email had to repeat the whole record. If a receiver missed the first message, revision 2 still contained a usable description. If revision 1 arrived late, its smaller number exposed the ordering error. This rule cost extra bytes to avoid making every library rebuild a current record from fragments of a possibly incomplete mailbox.

Withdrawal tested the same mechanism. The publisher could send another revision, keep the fields needed to identify the report, explain the withdrawal in NOTES, and leave TITLE empty or absent. RFC 1357 supplies that form. By 1995, RFC 1807 had added an explicit WITHDRAW field and still said that the receiving institution decided how to process withdrawn records. A catalog could retain an audit trail, suppress the old title from ordinary search, or follow a different local practice. The format made the instruction clear enough to handle; the service running the index had to handle it.

There was a more mundane way to pollute an index: test mail. The format marked versions beginning with X as experimental. Publisher symbols DUMMY, TEST and those beginning with X were not for permanent databases. That choice gave an importer a cheap filter at the door. A service that ignored it could make rehearsal records appear alongside research. Printable ASCII, 79-character lines, repeated fields such as AUTHOR and ordinary continuation lines kept the door usable for unlike mail and catalog systems.

Finding the card was still different from getting the report. RETRIEVAL could point to FTP or the post. If the address changed, a new bibliographic record could repair the pointer; the library then had to replace its stored version, and the reader still had to find a working copy. RFC 1807's later additions, including HANDLE and OTHER_ACCESS, show how much attention the path from citation to document continued to demand.

The 1992 email solved delivery of a publisher's latest statement. Cornell Dienst, Stanford SIFT and the later withdrawal field show why that was useful to real cataloging systems. A search result changed only after its keeper made the new statement part of the local record. The small protocol separated a fast announcement from the slower act that gave it public force.

Sources: RFC 1357, A Format for E-mailing Bibliographic Records; RFC 1807, A Format for Bibliographic Records; RFC Editor record for RFC 1357.