Separate group to discuss translation features

Also see http://groups.diigo.com/group/web-annotation-services

canonical example, threaded conversations, any entry in a threaded conversation should be able to be simultaneously a comment in that thread and the head of its own thread (a topic) these two 'types' could have both different interfaces and different affordances (the comment is just text in context, with associated order time author metadata, the topic may have subscribers, rss feed etc - a party to the original threaded conversation may ONLY be interested in that comment and its sub threads, and may wish to elevate this comment to a topic for example and invite others to discuss it - see perspectives also

using system 'Concepts' and user/group/social/domain specific types, and or not operations, result set type specification

take content on a page or in a selection and transform it to structured data, using system or user / group / social specified entities

recognize types such as books vids etc automatically fill the metadata aspect associated with them, allow creation of new types and type recognition training to system, a list type (any repeating structure on a page) could recognize list subtyp (books) or create / use a type for the underlying data (if its in a repeating list, it is structured, all it needs is a name) then feed it through a dabbledb type transform into structured database form, allow for social typing of data

some good stuff in here, starting around the 17 minute mark

perhaps not a fully fledged feature, but worthy of note, two semantic principles at work, one is more than one name for a given concept, the other is deriving a class from colocation of keywords or tags, probabalistic reasoning over a graph

identity management (federation, multiple identies, context specific sharing/privacy, authorizing to people/groups, commutation of trust)

note that many feel this should be a third party provided feature, at the very least it should have a layer of indirection between it and the remainder of the system

any graph is dependent on the user(s) and potentially the context, any user should be able to toggle view from private to shared (various groupings) to public views, (of course public and shared views require algorithms for composition

place time person etc, first degree concepts for faceted search, (these may have display formats (person shows avatar, place shows map etc), metadata aspects (lat/long, gmt+/-, name) search result display formats (timeline, map w points, social graph) etc associated with them)

extensible with:
domain specific ontologies, 2nd degree concepts
meta objects, 3rd degree concepts (non relational classification with metadata aspect attached, rich proto ontologies in formation)
defined tags (tags with defined meanings, loosely semantic tags)
dumb tags (grouping functions)

ambient location based presence (location may be physical or virtual)
promote concurrence

must have snappy performance, facets are user selectable from aspects and metadata, system should have a default set

given a node of information, it has a type (video, book, essay, microblog, wave) which has one or more possibly user specific layouts associated with it