Compression vs. Block

[cabo@…]

Klaus Hartke notes:

The Content-Type option includes the specification of a Content-Encoding such as "deflate" or "gzip". When used with block, should each block be compressed individually or transfers each block a part of the compressed representation?

If the former, which flush mode should be used, Z_SYNC_FLUSH, Z_PARTIAL_FLUSH or Z_FULL_FLUSH?

Carsten's point of view:

The intention is to transfer the content-coded representation, which is sliced up in blocks.
The sender (side that does the slicing) may want to do the compression incrementally, however, it has to make sure the slices are an exact power of two number of bytes. This makes flushing precisely at the block boundaries hard. (The sender may still want to do flushing, but this is an implementation decision.) The recipient obtains a content-coded representation that can be decompressed incrementally, but where the block boundaries do not mean anything and therefore don't need to be reproduced by flushing.

This item would benefit from input from implementers that use content-coding with CoAP.

[cabo@…]

Assuming that there is no further discussion, I'll reflect this clarification in clap-11.

[cabo@…]

Fixed in [1265]:

Fix #203
Fix #206
Fix #209
Fix #210
Fix #245