[OSGeo-Discuss] Open File Formats and Proprietary Algorithms

[Christopher Schmidt]

On Thu, Aug 20, 2009 at 01:57:16PM -0700, Landon Blake wrote:
> MPG:
> 
> Thanks for the clarification. 
> 
> When you said "there is today no open source implementation of JP2 that
> is suitable for geo work" do you mean that there is no open source
> library that can read and write JP2? If so, who is using the format?

There are:
 1. Several non-open source implementations (most of which cost money)
    which work at geo-sized JP2 images.
 2. Many use cases of JPEG2000 which involve imagery at sizes that are 
    less than geo. (This is the much more common case, in my research.)

> Do you know why there hasn't been a broader adoption of JP2?

I'm not sure what your definition is of "broader adoption"; many of the
datasources I worked with for OAM were provided in either JP2 or MrSID
formats. I would almost always go with MrSID, because I could:

 * Work with it easily, and for free
 * It was typically significantly smaller.

Perhaps you're asking why there hasn't been more open source software
written to handle large, highly compressed JP2 images better -- to which
I would point out that there isn't *any* format that has good open
source support for large, highly compressed images. (gzipped TIFFs work
to some extent, but don't compare to the benefits gained by JP2 or MrSID
in many cases.) It's a hard problem, and -- given that the major players
see the costs to 'pay to play' as being trivial (and they typically are,
in the big scheme of things), not in a situation where it's likely that
the people with ots of money ar ein a position to spend it on open
source, rather than simply paying a smaller amount for existing
non-opensource solutions.

Despite the claims that 'disks are cheap and bandwidth is free', many
providers *are* limited by bandwidth: MassGIS, for example, had to put
in cash for a costly upgrade to their badnwidth solely due to the demand
put on their servers by people downloading aerial imagery. Those funds
could have gone to funding more open geodata, but instead were used to
maek the data that already existed more readily available.

These things *do* matter, and MrSID offers, by far, the best 'bang for
the buck' for amount of data per byte of download. This applies even
more at the consumer end; when you talk about consuming data, MrSID is
even *more* user-friendly, because the users (who have limited
bandwidth) are able to open it more easily. Additionally, many viewers
which include MrSID support are able to display larger images -- due to
the MrSID library -- than they would be by opening the entire image in
RAM or something similar. Many of my friends have used MrSID for looking
at thigns like Shakespear's Folios, because tools like IfranView include
it by default, and the tool "Just Works" better than anything else.

I believe that the important things in terms of delivering public
content to users are:
 * License -- Are they allowed to do what they want with it?
 * Ease of use -- Is it *possible* For them to do what they want with
   it, including downloading it in the first place?
 * Openness -- Can they do what htey want with it with free/open tools?

If the formwer two are true, then the latter -- openness -- can be
handled by third parties.

Imagine that you have two options:
 * Data provided online, for users to download, in MrSID
 * Data provided on CDs, for users to have shipped to them, in GeoTIFF

(The latter will almost always have a non-trivial fee, because it
involves person time, but ignore that for the time being.)

If these are your options -- and this *is* the case for a non-zero
number of imagery providers -- which one would you prefer to use?

Best Regards,
-- 
Christopher Schmidt
Web Developer