Permalink
Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
82 lines (56 sloc) 2.94 KB
;;; :FILE-CREATED <Timestamp: #{2011-09-04T13:05:03-04:00Z}#{11357} - by MON>
;;; :FILE vivace-graph-v2-FORK/TODO
;;; ==============================
Following TODO adapted from the following two threads posted to
vivace-graph-devel mailing list:
[vivace-graph-devel] vivace-graph todo list
Kevin Raison raison at chatsubo.net
Tue Feb 1 00:15:23 EST 2011
(URL `http://lists.common-lisp.net/pipermail/vivace-graph-devel/2011-February/000000.html')
[vivace-graph-devel] vivace-graph moving again
Kevin Raison raison at chatsubo.net
Sat Jul 2 17:00:43 PDT 2011
(URL `http://lists.common-lisp.net/pipermail/vivace-graph-devel/2011-July/000002.html')
I (mon-key) have replaced the original per item numeric annotations with
asterisks and changed the read order to accomodate re-grouping of related
tasks. This re-grouping occured under the assumption that the original numeric
ordering of KR's requirements did necessarily imply that some requierment-A was
implicitly more important / required than some other requirement-Z.
;;; ==============================
The major things that need to be done to satisfy my personal requirements:
* Stabilize transaction logging.
* Remove the SBCL-specific code that litters the project.
* Make a choice on what sort of on-disk storage mechanism to use.
Currently, all data must fit in RAM and is persisted via snapshotting and
transaction logging.
I would like to add a second option where, as in a standard database, data
can be stored in a disk-based structure.
- B+Trees
- Fractal Prefetching B+Trees
(http://reports-archive.adm.cs.cmu.edu/anon/2002/CMU-CS-02-115.pdf)?
- B-Tries
I expressed concern to Red and Ilya about using B-trees, given the
extremely high fanout of the indices.
I was considering something like B-tries, as described in this paper:
(URL `http://www.naskitis.com/naskitis-vldbj09.pdf').
* Spend some time on the text indexing piece.
It currently uses my cl-skip-list, but should be moved into a B-tree.
* While triples are thread safe, Prolog may or may not be.
This should be investigated and fixed where necessary.
* Fix lexical env issues between Prolog and Lisp.
Currently, importing a Lisp variable into Prolog requires that it be
declared SPECIAL. This is because the Prolog implementation is based on
Norvig's PAIP, which uses CL:EVAL. A better solution should be found.
* Add reciprocal reasoning.
(has-child implies has-parent)
* Factor out the certainty factors code.
- Make CFs first class triples
Currently these are a mutable fields of the structure TRIPLE.
* Incorporate Geospatial / Temporal reasoning
* Incorporate forward chaining rules engine.
(Rete or LEAPS-derived)
* Incorporate more RDF-related stuff.
(Sparql, N-Triples parser / loader, etc.)
* Implement a protocol for submitting queries over the network.
;;; ==============================
;;; EOF