Thoughts on Longwell

From: Ben Hyde <bhyde_at_pobox.com>
Date: Fri, 10 Mar 2006 10:38:34 -0500

Some thoughts moves we might make in the longwell design space.

David Huynh's suggested that some simple editting might be easy to
add at this point.

Switchable item displays - The individual items are currently shown
thru a fresnel lens if available, and then fall back on a default
(primitive) display. I think switchable displays would be great.
That would let us have fast summary views of things and really
complex views the user can play his way into.

Revealing dark matter - Stefano used the fun term dark matter to
describe how longwell, particularly now that we have these beautiful
lenses, can leave swaths of the data invisible to the user.
Meanwhile I'm observing that each time we get a clear view of another
fragment of the converted data sets we discover bugs. It would be
nice to get to the primitive display close at hand all the time. We
are far too blind to what the underlying data is and given how
recently it was converted that means we don't see bugs.

Non-fresnel lens - There are obviously plenty of other flavors of
item display that would be nice to have at hand (n3, those silly
little graph displays, etc. etc.). My primary motivation here is
debugging though.

Lazy facet modeling - Could the tiles in the facet display be brought
up lazy, on demand? We only need a full model of the facet after
it's tile is unfolded. My model of how things work suggests this
would be a huge performance win. I think it's tolerable to show
folded facet tiles which once opened and you build the model for that
facet you discover - opps none of the items in the current query
result actually have that facet.

Sort order control - Adding a control in the filter summary (i.e. the
query) that allows users to control the sort order would be nice.

High speed debug mode - If you have lazy facet modeling, sort order
control, and primitive or n3 displays at hand then you could spin the
dials to turn off sorting, and defer all the facet modeling. This
would have three positive consequences. It would make it clear where
the performance sinks are. It would allow us clear vision of what we
are debugging. It would allow this tool to be used much earlier in
the data import pipelines.

Gadget like facet browsing - This is obvious but I wanted to mention
it; particularly the spark lines. :-)

  - ben
Received on Fri Mar 10 2006 - 15:38:33 EST