WiredTiger's file format doesn't note the compression algorithm that was used anywhere in the page, which means applications cannot use multiple compression algorithms in the same application (pretty unlikely, I suspect), or change from one compression algorithm to another between releases, without a full dump/load cycle (yeah, that could happen).

We could use the WT_PAGE_HEADER.unused field for this purpose, but there's no obvious name mapping for each configured compressor.

We also have built-in support for bzip2 and snappy, so we'll need to split the name-space somehow so we can give the built-in compressors their own IDs without colliding with application compressor extensions.

Finally, there's no way to specify multiple compression engines for a single file: the compression engine's name is stored in the metadata file, but there's no method of replacing that value. (I suspect that's all we'd need: we could use the compression engine ID to identify the compression engine to use for decompression, and the named compression engine would be used to compress all newly created pages.)