ian@minster.UUCP (ian) (12/12/85)
Suppose someone at institution X has just completed a technical report of interest to you. How are you going to find out about it? How long will this take? The purpose of this article is to propose a new solution to this problem. I'd like the originators of technical reports (authors or institutions) to post references to new reports to the net. No, I don't expect people to read all of them! Rather, I envisage selection of references by keyword search. Some sites might collect references to aid searches of old literature. This isn't so far-fetched - York already has a system to do most of this, but it relies on local people (who've all got other things to do) typing in the references. Our system would provide more timely information, and would involve less work for us, if the references were to fall off the net. Surely the same situation must exist at many other institutions. Note that I'm thinking of lists of references, not annotated bibliographies. Keywords would be encouraged; abstracts probably discouraged (due to volume). I haven't mentioned journals due to the organization needed to avoid duplicate postings of the same references. (Maybe if publishers were on the net...) Postings of references should be confined to a separate newsgroup, both to avoid cluttering other groups and to ease automation. A subgroup of net.research seems most appropriate. Maybe we need separate newsgroups for different research areas - we're not all computer scientists. That's the basic idea. Perhaps someone else would like to comment on the "Data Protection" implications of this proposal. (Yeah, the UK is quibbling about protecting bibliographies while excluding many fishier uses of information from any legislation whatsoever.) I admit that we've all got more than enough to read already (never mind news), but this tool would improve the quality and timeliness of our reading matter. So, how about net.research.cs-refs? Please *mail* opinions to me at the address below, and I'll post a summary here. Do *not* send votes. The rest of this article describes the York references system in more detail; the news gurus out there might like to ponder what else would be required. (Don't expect me to volunteer to implement my suggestions - I have a DPhil (=PhD) to complete and time is running short.) All our references are in pubindex(1) format. Obviously a standard format for posting would be required, so I would propose pubindex format. The organization of our references system is unsophisticated. Every night, cron sets off a process which searches for files that have been modified the previous day and which contain references. These references are collected together into a single file, which is then hashed using the standard UNIX pubindex tool. Next, each users' home directory is searched for a file called .keywords; relevant references are selected from the new ones by lookbib, and these are mailed to the user. This process also rebuilds an index of all references on the system (if necessary). We have separate tools for retrieving references from this collection (a simple shell script around lookbib) and a screen-based browser to aid the viewing of references. mail replies to ukc!minster.york.ac.uk!ian-refs