Ontaria Development Roadmap

This document is a high-level development plan for the Ontaria service and the underlying SemWalker code. It is subject to change, of course; this is not a committment to deliver these features or meet these target dates. On the other hand, this is our best guess at how we will proceed. We welcome feedback and even assistance. This is all open-source.

Releases are numbered in a very common way: <major>.<minor>.<patch>[<qualifier>]. Each minor release brings a new set of features; each patch release fixes bugs. Qualified releases (such as a release candidates like 0.8.1rc2) come before unqualified releases, and should not be considered stable.

Nearby, the SemWalker TODO file is a place for developers to keep related notes from a developer's perspective, and the Ontaria Use Cases (partial draft) further describes the problem-space.

Sometimes pre-release versions are running at the Alpha and Beta development/test sites.

Release 0.7 (focus: up and running)

Suitable for private demos

• Stable platform: multi-threaded HTTP server, available to the public. Downtime monitored.
• <1 sec response time on most operations with static database load of >10K real-world ontology triples.
• Text (keyword) searching for objects
• Browseable property-table views of RDF resources: all arcs in and out displayed in a table, with arcs and nodes labeled and usable as hypertext links to a new table focussed on that resource.
• Multiple views of a resource, when available, using a tab metaphor
• Simple, brief help text
• Visually pleasing and effective layout and styling, on at least mozilla firefox. (No need to work around CSS bugs in other browsers yet.)
• "Add a Site" page, for users to submit new ontologies/data sources
• Simple front page: Ten-random-subject, Ten-random-sources, Simple triplestore stats
• Cross reference term use, showing (and counting) where each term (especially class and property names) are found to be used.

Deployment Target: 25 May 2004 (Done)

Release 0.7.1 was shown by Ralph Swick to the DAML PI meeting on May 25. Feedback included requests for a feedback page linked from every page, quality metrics on ontologies, and links for each ontology to open-issues/best-practices data.

Release 0.8 (focus: harvesting, scaling)

• Real-Time Harvesting. RDF content loaded if possible from URIs used in browsing. Web content cached for performance and archived for look-back on historical content. [maybe 0.8.1: HTML metadata loaded (header links to RDF followed, to give users feedback on HTML deref; Expiration handling; Manual Refresh option].
• Precache. Have a base dataset of few hundred useful RDF pages / ontologies, including those explicitely submitted by users, and those collected by the other sites, such as the DAML Crawler, DAML Ontology Registry, and SchemaWeb.
• Sub-second response on typical page generation using < 1 million triples.
• Sub-second response on keyword searches over < 1 million triples.
• Draft QuickStarts (Selecting Ontologies, Using Ontologies, Creating Ontologies, Publishing Ontologies, Giving Useful Feedback)
• Simple provenance: List of RDF sources used to present each view given. (But sources are not connected to triples for the users yet.)
• Client-side regression testing to help assure site quality, availability, and performance.
• Export internal data (eg cache metadata) on the web as RDF/XML.
• Prototype rules for inferring which RDF documents should be considered to be Ontologies.

0.8.0b1 installed on Ontaria Beta, 17 June 2004.

0.8.0b2 installed on Ontaria Beta, 28 June 2004.

Original Deployment Target: 1 July 2004. Interrupted by Ontology Hosting Project, taking 2-3 months developer time.

Deployment Target: Friday, 24 September 2004 (Done)

Release 0.8 was announced on 23 September 2004 to the Semantic Web Interest Group.

Release 0.9 (focus: ontology metadata and presentation)

• Ontology Hosting
• Gather and report usage statistics
• Support HTML inside RDF content (so ontology documentation can use hypertext)
• Domain/Range/Inheritance chart (on a tab where applicable), to simplify visualization of the relationships between classes and properties.
• Single-page view of an Ontology, suitable for using as the namespace document of a HashURI-style ontology.
• Some inferences done, including standard ones (some of RDFS and OWL).

Current Deployment Target: 28 October 2004

Release 1.0 (focus: ready for new users, lots of users)

• Neighbor dereferencing (aka crawling, harvesting), including robots.txt handling. Requires very robust handling of bad data.
• GRRDL support, for getting RDF data out of HTML.
• Per-triple provenance (on at least the Property Table tab), even for triples derived by inference (where provenance is a proof step).
• Usability Testing
• Category-Based Ontology Browser. Like Yahoo! or the open directory project, for ontologies. Perhaps a front-page box like dmoz suggests (scroll down).
• Proceedures in place to limit service downtime due to system failures or surge loads (slashdot effect).

Other important features/changes are likely to emerge by this point, especially based on feedback from the 0.9 set of ontology presentation features.

Current Deployment Target: 2 December 2004

The "Someday" Pile (post 1.0)

Several of these involve Ontaria having a notion of user identity.

• Make the Ontaria/SemWalker split more serious, eg in the documentation and static files.
• Shopping Cart. A way for users to indicate which items are of special interest to them, perhaps for aggregation and republication as a new ontology.
• On-Site Authoring. Fix typos immediately, when you see them. Or even add new terms, descriptions, etc.
• Trust Profiles. There should be different sets of rules saying how much to trust each source; users should be able to create them and select which ones they want to use.
• FOAF-specific views, for navigating social networks
• Handling images. Collections of images, and nearby images (eg foaf:depicts)
• Downloadable views: view-rules coming from the web.
• Presentation of User Reports (of ontology usage) found on the web (or in mailing list archives); Ontology rating (in some inoffensive way, supportive of incremental development) based on these.
• Front page awareness boxes: recent user reports, most used ontologies, best-liked ontologies.
• Feedback loop. Ask for feedback (eg BugReport and RequestsForEnhancement messages; allow them to be browsed on the site; explore the overlap between user metadata about software systems (especially Ontaria) and ontologies. May obsolete this roadmap, or automate its generation.
• Validation: present users with some automated checks on ontology correctness, using an external validation program or service.

[For staff use: editable version.]