Hi all
I'm going to keep it short (ok, scratch that, actually not short!) and simple,
first because I don't have the excess brain power to really consider these
proposals fully :) and I'm busy here in Kenya, but mainly because my feeling is
that it's important for OAM to have a gradual ramp up curve, and tangible
roadmap. It's great that there are resources at NPS, which we should try to
take advantage of, but as pointed out before OAM needs to be robust against
single points. We're a relatively small group, distributed, with limited time
and resources. In order to get the ball rolling on OAM, I challenge you all to
ponder "what is the simplest thing that would work and incorporate some idea of
OAM?".
I've been involved in OSM for four years, and in the early days it was often
just 1 or 2 people actually building something. It took very little resources
to build a vector data commons, when the data contributions were from just a
few folks. The basics of the API was set relatively early, allowing many tools
writers to come in and out of the proejct, while the actual server software has
been rewritten many times as OSM has dealt with massive growth. If OSM had
attempted to deal with the TIGER import back in 2005, it would have doomed the
entire project .. and that's essentially the challenge OAM has taken on up to
now.
So I like Tiles. Tiles are a simple "standard" already widely supported by
clients and processers, and one aim of OAM should be to catalog these
techniques, promote best practices etc. For instance, see Josh's tutorial on
tiling our Kibera imagery
[http://porcupinealley.com/entries/2009/oct/15/tiling-kibera/]. This is
potentially understandable to someone with little background in GIS, or open
source .. and one potential user story for OAM could be an overworked,
non-specialist IT guy in a small municipality, who's been delivered a GeoTIFF
from an aerial survey .. how can we quickly help him get that imagery into OAM?
I caution against limiting the number of layers. This should be unlimited. It
could even extend beyond aerial imagery. And I also caution on attempting a
global mosaic from the start. It's very hard, and yet without a global mosaic,
with a simple system for distributing tiles, there would be a very powerful
first step towards ramping up OAM. The simplest thing I can imagine is just a
catalog of tile sets, with some metadata, bbox, and a simple API for clients to
search that catalog. Anyone could register a tile set with OAM, and for
example, and OpenLayers extension could be written to interact with that
catalog. The tiles don't even need to be cached to start .. just make requests
direct to the provider. As the issues with caching and distribution get worked
out, those systems get integrated into the existing OAM.
On governance, something like what Schuyler describes seems like a good deal.
We need some mechanism to move. I guess we can learn from OSGeo here. The same
with documentation of this conversation. I second Charles' call to bring
discussion into the wiki, but realistically, someone needs to step up to focus
on this.
My thoughts .. thanks all for keeping the OAM dream alive.
Mikel
_______________________________________________
talk mailing list
[email protected]
http://openaerialmap.org/mailman/listinfo/talk_openaerialmap.org
