[OSGeo-Standards] FOSS in the OGC mass market

[Jo Walsh]

dear all,

Here's a writeup of the short talk I got roped into giving during the
"Mass Market" session at the last OGC TC on the "Open Source Perspective"
It was a last minute thing, thus scattershot rather than strong message.
I would usually "blog" this sort of thing, but the places i normally
write don't really fit it. Perhaps there is some shared blog space on the
OSGeo.org site if deemed suitable. Pls someone yelp at me if i am saying
wrong things about WMTS, etc.  

----------------------------------------------

I was able to give this talk at all because of OGC's generous offer of 
10 individual memberships to OSGeo members in an attempt to bring more
engagement from free and open source software projects into their 
process of standards creation. However a request for an "Open Source
Perspective" on OGC and standards bodies at large is a tall order,
because the free and open source movement in geospatial software and 
data is too copious and broad to talk about one "perspective".

Within geospatial software, there is a spectrum of approaches to standards.
Near one end is the GeoAPI project (http://geoapi.sourceforge.net/), set up
specifically to stresstest via implementation, ISO and OGC standards 
for geographic information modelling, figure out where the holes are
and attempt to fill them. Now the maintainers are looking to set up an 
OGC Standards Working Group to formalise GeoAPI itself as a specification;
allowing the easier migration of components, common testing setups, etc.

Far along the spectrum is the open source / open data project OpenStreetmap,
which has taken a conscious decision to create interfaces independent
of "standards", by developer fiat or by the slow accretion of feedback
from members of the user community into an ever-evolving de facto 
standard with no specification. For the project's founder, Steve Coast,
"Real artists ship. Everything else is standards-wanking."

The projects that form the OSGeo foundation are clustered at the 
"standards-facing" end of the spectrum, having benefited a lot from 
support for the OGC Web Services standards, particularly the Java
projects that have been involved in the OGC funded "testbeds" to create
reference implementations that become a kind of "gold standard".
In this sense, OSGeo *is not* the Mass Market that you are looking for,
but it can provide a bridge to it, especially through easy-to-use 
web-facing client software packages such as OpenLayers.

I'm still trying to make sense of the OGC's "Mass Market" activity
as a continuous project. The efforts that have wound up inside it -
GeoRSS and KML being the popular successes - suggest some commonalities.
The Mass Market efforts seem to be driven by *implementation* - de facto
standards common in the wild, having a formal specification created for 
them. They also seem to be somewhat *self-standing* relative to the main 
core of OGC standardsm, which must (in theory) comply to its Abstract 
Specification. The Mass Market specifications are correspondingly *simple* 
- to understand and to implement, without the time/energy investment i
incurred by the full OGC and ISO abstractions.

So for efforts possessing these three qualities, OGC is offering a
different kind of process, through its Mass Market effort. This involves
access to drafts earlier on in the spec creation process and more 
opportunity for public consultation (such as that going on right now 
over the KML 2.2 spec and schueduled to end on Jan 4th 2008:
http://www.opengeospatial.org/standards/requests/45 ) 

More open inspection earlier into the process, incorporating feedback
from anyone in a position to give it, is a a behaviour that people 
working on free and open source software take for granted. 
More open inspection can only provide better strength in standards 
and wider implementation.

One particular specification I've been asked to mention here is that
for tiling of map data delivered over the web. There have been arguably
three de facto standards common in free software for web map tiling,
efforts going back a year and a half at least. The implementors 
responsible - particularly after a broadly attended meeting at the FOSS4G
2006 conference - have had no chance to participate, even in consultation,
in the WMTS efforts happening inside the OGC, which don't look much like
the existing free software prior art. For implementors this isn't 
necessarily a problem - a piece of software like TileCache will support
*any* XML-based standard for transferring tiles around, offering a common
software API for reassembling them, so one more in the mix isn't a problem -
but it is troubling that there's been no means to fix or strengthen 
what will become "industry standard" until so late in the process. 
Treating efforts like WMTS as an OGC Mass Market project would help engage
the knowledge and energy available in free and open source "communities".

Our biggest problem is time. Although some people are supported by 
corporate entities to work full or part time on open source software,
the majority of development activity is a byproduct of project and 
consultancy work. What may be an issue of pressing relevance now, 
will have been assigned to the back-burner in three weeks; if spare 
developer time is available and it's not taken, it's likely wasted. 
Within the OSGeo foundation many people are donating their time *because*
they're already donating their time to other projects; while the 
commercial benefits of being directly involved in the specification
development process are *no less*, people in GFOSS are much less likely
to be able to afford the time investment upfront, that people who 
are supported specifically to be in the standards process for business
and/or governance reasons. More information, earlier, is the core answer.

Other "open issues" that I've heard raised amongst FOSS4G developers 
are really to do with *very* simple problems, perhaps so much so 
that they fall below the radar of standards-body participants.
>From software projects and infrastructure designers I've heard 
pleas for some common way - using a URL or URN scheme - to refer
to different kinds of, and versions of, web services for delivering 
geographic information, that isn't specific to OGC and can be easily 
extended. Everyone winds up using their own custom scheme, there's 
no simple common mapping. OGC has a URN standard for identifying 
web services, which other efforts built on, but it doesn't extend
to describe KML, or ESRI-specific, or other interfaces.

Another issue is a similar one, common schemes for referring to 
and describing coordinate reference systems and projections,
which extend further than the well-used EPSG codes, which are by no
means copious. I was surprised to hear excitement about the development
of common libraries and descriptions for CRS/SRS specifications at 
the last FOSS4G - it seems like something so fundamental to all the
work with geographic data that we are doing! If the OGC is in a 
position to address and fix these simple issues, it would be doing
a great service to everyone working with geospatial software and data.

As OGC participants have seen over the years, getting their standards
embedded in free and open source software encourages "acceptance", 
helps partial specifications to harden and mature, and promotes
widespread adoption. Increased engagement, given the right balance of 
time/money/energy, can only be a good thing.

--