Purpose of Supercompressed Global Data block about ktx-specification HOT 5 CLOSED

Crunch CRN has a code book common to all mip levels. Before we can inflate any mip level we must have the global code book. There are 2 ways to accomplish this: put it at the file format level, or specify that the first mip level transmitted must contain the code book. The first seemed more reliable and easier to specify since we are supporting random access to the mip levels and the ordering of the mip levels can change. But it is not a strong preference.

Apps adding their own global data is already supported. It's called metadata.

from ktx-specification.

Another reason for having the global code book at the file format level is it is less likely to be missed by those using, e.g., HTTP, to read ranges from the file. If it is hidden in the first mip level, a user may directly read a later miplevel then discover they can't inflate it.

from ktx-specification.

Understood. I would like to discuss this one some more, but it is not a blocking issue - the data block should be supported, its just clarification on the various things we'd need to use it for

from ktx-specification.

On the 12/21 call we reached consensus that the supercompressed global data should not be made available at the file level, given that the data should be abstracted in the decoder. The codebook binary block will instead be relocated to be the first data block immediately prior to the highest level mip level in the data stream.

This issue will be closed, and a new one opened to track the work item

from ktx-specification.

This is not my understanding. We agreed to move the block but it would still be visible at the file level due to wanting random access to the mip levels. Let's continue further discussion on issue #39 you opened to track to work.

from ktx-specification.