On May 30, 2007, at 11:34 PM, Guy Rixon wrote:
> Supporting Rob's point, we _should_ use image/fits for single
> images and we _should_ have a place in our protocols to state and
> control content encoding that is separate from MIME type. Then the
> problem goes away except for choosing out a token that means "tile
> compressed". Until we put this into our protocols (or until the
> FITS committees bless tile compression, or the heat death of the
> universe, whichever comes first), our services _should not_ be
> emitting compressed images.
Um...this statement seems a bit strong. Tile compressed FITS files are completely legal FITS, worthy (at least) of an application/fits imprimatur (as with lots of other useful bintable applications).
Data compression is a key technology for many purposes. The statistic from my ADASS talk was that the additional cost of 32-bit + gzip compression was $2.86 - per image - more than the cost of 16-bit + Rice tile compression for each Mosaic camera exposure input into the NOAO pipeline and archive. NOAO has > 10^5 Mosaic images on hand. A quarter million simoleons is real money even at today's depressed exchange rates.
This is rather far afield of SIAP, but similar considerations will apply for applications layered on VO services.
Rob
im•pri•ma•tur: n. Official approval or license to print or publish, especially under conditions of censorship. Received on 2007-05-31Z09:10:34