[postgis-users] Proposed SQL interface for PGRaster

Stephen Marshall smarshall at wsi.com
Tue Dec 5 13:36:14 PST 2006


Mark,

I agree that the user needs to know about the endian-ness of the data.  
However, I'm envisioning a system where the data is returned in a 
well-known image format, e.g. Geo-TIFF, JPEG2000, etc.  Thus I'm 
delegating the requirement to the storage format.  These formats should 
contain either conventions about endian-ness, or a parameter that 
describe which endian-ness was used. 

I think your concern could be addressed by adding an additional 
requirement for an acceptable raster format:
* Byte order of the data values, e.g. big endian or little endian

Do you agree?
Steve

>
>Hi Steve,
>
>Thanks for an interesting proposal. My only comment so far would concern
>the endian-ness of the raster data if it is stored as binary - for
>example, what would happen in the case of dumping a PostgreSQL database
>containing pgrasters on an Intel machine and restoring it on a Power
>machine with respect to the int16/32/64 and float32/float64 types?
>Perhaps you need to store the endian-ness of the data as it is inserted
>into the database, and then perform optional on-the-fly conversion
>depending upon the current processor?
>
>
>Kind regards,
>
>Mark.
>  
>




More information about the postgis-users mailing list