Email Fields, an RDF Schema (original) (raw)
A Knowledge Base about Internet Mail
Recent (Feb2002) development:
- extracted 18MB of RDF/xml about 16000 messages using <mid%5Fproxy.py>, which proxies IMAP queries over HTTP, with results in RDF (or HTML). (it started as a proxy resolver for mid: URIs)
- ran some rules over all that RDF. (the computation used over 200MB of RAM; it finished in somwhere less than 6 hours on this 1.7GHz/1GB machine). The rules are, roughly:
- Replies come from people; if a message has an in-reply-to header, then the address in the From header is the contact:mailbox of somebody, and the phrase used in the From header is the contact:fullName of that person.
- If I wrote to the mailbox of a person, then that person is a contact of mine.
See the for details.
Next steps:
- enhance toAddrBk.py, which produces ldif suitable for import into evolution, to take an RDF KB as input.
- think a bit more about the schema for mail:to and such; publish it.
See also Internet mail message header format by D. J. Bernstein for lots of practical info about parsing email. I implemented some of it in a perl script.
Other mail-messages-in-the-web thingies:
- metamail 0.1 An experimental RDf Vocabulary for describing mailing list messages ; discussion 5Sep2001 with filsa... brother vs brothers and such. (chump from 5Sep has lots of goodies)
- IMAP. See the IMAP spec (html), my<imap%5Fsort.py> hack, python imaplib doc, my <imaplib2.py> revision.
- Eyebrowse looks cool. cf my 18 Aug 2000 suggestion that they consider RDFDB.
- Gerald's HURL looks promising too. (also: Hypermess, mailman, @@hypermail done right thread on www-talk)
- A MIME->XML thing: XMTP: XML MIME Transformation Protocol Copyright (c) 1998-2000 Jonathan Borden ALL RIGHTS RESERVED; hm... the namespace name,
http://jabr.ne.mediaone.net/xmtpisn't very useful for finding out about it. Sigh.
@@compare/contrast with earlier stuff, e.g. foaf:mbox , foaf:name, rdfs:label.
See RDF and the RDF interest group archives for more context.
Phrase
the phrase part of a mailbox, e.g. Dan Connolly
@@say that this is a subproperty of rdfs:label? (as are foaf:name and dc:title?)
Date
date(mid:222, "Mon, 22 Jun 1999 12:33:23 CDT") ::= mid:222 bears date 'xyz'. see RFC822, section @@
note that case matters in XML and RDF, while it doesn't in RFC822 header field names. So this one property corresponds to all case variants (Date, date, DATE) of the field name.
From
To
note: sometimes it's just mid:foo--to-->mailto:bar, but sometimes it's mid:foo--to-->[recipient]--addr-spec-->mailto:bar . Same applies to From/CC as well. @Hmm... maybe I should use different properties, related by a rule.
Hmm... maybe the object of the to property should be a list so that we can record that these were the (stated) recipients, and there were no others. I say stated because there could be blind copies.
Carbon Copy (cc)
Dan Connolly
Revision:1.10Revision: 1.10 Revision:1.10 of Date:2002/02/1223🔞39Date: 2002/02/12 23🔞39 Date:2002/02/1223🔞39 by Author:connollyAuthor: connolly Author:connolly