Does it make sense to support a binary encoding?

[fdr]

I've noticed that encrypted payloads are base64 coded. Does it make any sense to have a binary mode? I'd like to use Fernet in a high-volume way (maybe that's against design ethos, I am not sure) and avoid paying for base64 in size blowup if possible.

[kr]

Yes, totally makes sense IMO. I've wished for this myself once or twice.

[fdr]

How would you go about doing it while keeping Fernet's format self-describing? Use of the "version" byte (more like "variant" in this case?)

[kr]

My first thought is to have two separate entry points and require
the user to know a priori which one to use. As you note, the version
byte would not match and/or it would be an invalid base64 encoding
if one used the wrong one, so there's no danger there. But somehow
it feels like making the programmer explicitly choose is more better?
It's what I'd expect anyway.

[kr]

I guess that would be sort of against the "you can't do it wrong" spirit
of fernet though. Open to suggestions. cc @tmaher

[tmaher]

I'm weakly opposed to adding it as an option, mostly to prevent option bloat.

@alex - I'm more strongly opposed if pyca isn't going to add it as well.

[alex]

I'd have to think about it, but probably we'd add it as well /cc @dreid.

[dreid]

I'm +1 on supporting the pre-base64 encoded form.

I assume this would just be for tokens and not affect keys?

[fdr]

Yeah: tokens only.

[kr]

Though this has crossed my mind, I've never wanted it enough
to even bring it up in conversation before. And I agree with the
desire to avoid option bloat.

Note that one can do it oneself as a workaround. It's not hard
and IIUC not security sensitive to undo the base64 encoding
before storing or transmitting it, then redo it before decrypting.
It's wasteful of CPU, but if the point is to save space, maybe
that's okay.

[fdr]

How about this:

Before making any detailed decisions, I do have a specific application and lightweight benchmark. I'll hack in binary (aka "does less") fernet and see how much it matters, and post my result.

[kr]

@fdr any results from your performance tests?

[fdr]

On Tue, Sep 2, 2014 at 1:10 AM, Keith Rarick [email protected]
wrote:

@fdr https://github.com/fdr any results from your performance tests?

Regretfully I had to abandon the project before having the need to complete
them.

[rafael2k]

Hi all. I'm interested in using binary bitstream communication using Fernet. How is the status binary encoding support in Fernet?