August 20 conference call

Stefan, reporting on the hackfest:

Modular system, or a collection of loosely connected applications? Still hoping for the former. What to do with integration of Featurelens? Featurelens or Nora could initially be applications that work in the workbench framework, and we pull features out of them, one by one making them modules, until the applications themselves go away. Monkbench architectural metaphor, for now - the workbench has decks of cards, from which you can pull the deck or the card that suits your purpose. Start by using the whole deck, later use cards in different combinations. Flickr is another metaphor: drag and drop your modules into a workflow (photo organizer). Now, what about D2k? Like Yahoo pipes, constrains inputs and outputs that must be constrained, but on the other hand thinking of things in this way could focus conversations between data and analytics; perhaps, though, interface at this point should be thinking about drafting an interface for SEASR to actually build, so that we don't duplicate effort. Amit will talk to Duane about this. Unresolved issue: server-side or client-side? Server-side gives you strong typing; RAP (still in very early stages) allows us to develop in Eclipse, but for the Web - might be a good kit, but it's very young. The other approach is to do it client-side in some kind of javascript language, more weakly typed, very lightweight: advantage is that it allows you to build components very quickly, with very little component coordination, could be in any language, in any library - but this may not be robust enough for a full Monk environment. Amit should also talk about these issues with Duane....and the next interface call will revisit the issue (perhaps with Duane on it).

Proxy Server:

A key piece in the architecture, as a broker. In one example, it could broker between nora and 19th century text; in another (more modular as Stefan says) it might be doing more complicated work. Question for now: do we have a monk data-store that could be addressed via a proxy server from nora or wordhoard. Server-side (RAP) option might do a lot of that work, prior to the proxy, though - so let's focus on the proxy as a data-broker, between content system and application. Now, take the example of the 19th-century Text Collection integrate into WordHoard: can we now point Nora at this collection? Also important is getting the NCTC into a nora datastore, for purposes of comparison. Both are very important - perhaps Steve works on proxy to WordHoard, Amit works on ingesting, the purpose of both exercises will be to meet some specific use case at hand, e.g. Sarah's or Tanya's. Stan will review proposed functionalities/interfaces with them, at the same time, with an eye to what proxy calls and what data are required to support these (and where have we built something that looks like this, in WordHoard, Nora, Featurelens, etc.?). Proxy documentation on inputs and outputs and installation exists.

Chair of Analytics, Data:

Amit will hand over to Bill, Steve will hand over to Martin.