Jay Berkenbilt

> Hm, this doesn't work on mbedtls 2.28, I'll see if I can do something about it. Most likely, we would have a test in CI that would just pull...

I'm okay with it. I'm opinionated about CI, but if you get it reasonably close, I can rework it. There are some things about my human API that are undocumented....

Thanks for all your hard work. I would like to review this even if @m-holger also reviews it, and I'm kind of slow because I have limited qpdf time. I...

I added the `next` label in hopes of finding time before the next release review this and get it in/iterate on it. I want to get a release out with...

Re quotes, qpdf doesn't see the quotes at all. That is handled by the shell, which ruby is likely invoking depending on how you invoke qpdf. The quotes would matter...

> Was the intention perhaps to disable `preserve_encryption` if `compress_streams == false`, which feels more in line with the comment? Any thoughts? Most likely. stream_decode_level was added much later, and...

> My interpretation is that if `--compress-streams=n`' and `'--decoded-level=generalized`, Flate compressed streams should be uncompressed. Honestly, I get tangled up on this every time I need it. It's a bit...

Actually, on the original issue, I'm not sure we want to change the behavior. We might want instead to update the documentation and say that the default decode level is...

It might be worth double-checking what the behavior was before QPDFJob was introduced. I will want to review this PR before it goes in.

> I think the intent was to uncompress/recompress any streams where we know that we can improve on the existing compression without loss of information. That suggests that qpdf_dl_generalized is...