[Geotiff] Re: [UMN_MAPSERVER-USERS] BigTIFF Implementation Project - Seeking One Additional Sponsor

[Niles Ritter]

On Sun, 2006-12-24 at 09:29 -0500, Frank Warmerdam wrote:

> I just wanted everyone to know we have our fourth sponsor
> and BigTIFF is a go!
> 

Hi, congratulations Frank, et al. Just a few quick notes from what I've
read so far:

1) Are there any IP/patents filed by the original designers
   of the BigTIFF concept ?

2) The current BigTIFF design is a little bit away from the original
   "spirit of TIFF". Clearly once you get into the true
   64-bit address space the old TIFF readers will simply
   have to give up. However, setting the TIFFVersion=43
   value forces a reader to give up immediately, without
   even getting to a single tag that would say in text,
   "sorry, this image is so cool we won't tell you anything
   about it, so there."  

   Almost hostile, if you ask me.

   It would be cool if the BigTIFF design was friendly
   enough that it would at least permit old-school TIFF
   readers in the data pool long juuust enough to see a puny 
   little 4 gigabyte thumbnail of
   that 200 bongobyte big image of Paris Hilton's navel you
   got stored at offset = ten billion.

   The 8-byte offset to the real stuff could be stored in
   a reserved tag, whose presence will tell those in the
   know that they will need some big-iron software to
   read the whole thing.

   It was this same dream of "coolness" that prompted us
   to drop the "Private IFD" approach to the original GeoTIFF,
   in favor of using standard tags that would at least be
   visible to naive TIFF dump programs, so the user could
   get a clue what was in the file before the reader gave up.

3) Using an "overloaded" BigTiff would have the added
   plus of allowing the standard GeoTIFF tags to 
   provide some geolocation, at least of the thumbnail,
   even if you don't know anything about the big image.

4) A multi-scale lossless compression scheme may be good 
   for facilitating progressive resolution display of BigTiff
   (ala Google Earth's successive refinement of a
    satellite view over time). This might even be more
   of a data-structure option (e.g. strip vs. tiles vs. multiscale tile)
   rather than a compression scheme.

Just my two cents.

Good Luck!

--Dr. Niles Ritter (author, GeoTIFF 1.0 spec and libs)