What's encryption?

OpenSSL instead of PGP for message signing and encryption? Really?

• Over the last several years, I've seen more developers use openssl for file and message signing and encryption, where in the past we would have used PGP/gnupg/OpenPGP.  Is this wise, is it essentially throwing away the work that went into the PGP web of trust, and should we worry about it, or just go with the flow?   I'm guessing one reason for openssl's inroads into signing is that the APIs for PGP have been pretty slim (I was one of the earlier authors of python's 'gnupg.py' workaround; desperation leads to desperate code...) Anyone have any strong opinions about this?  Does the value of PGP's web of trust outweigh the implementation headaches, or is it time to move on?  Openssl is certainly a crypto swiss army knife, can do signing and file encryption just fine, and has better APIs, but frankly, the top-down SSL certificate infrastructure makes my skin crawl.  Now I'm at the point where I need to pick a crypto suite for a new open source peer-to-peer application.  The application is message-based rather than stream-based, will use websockets between peers for message framing at the network layer, and needs both wire and on-disk encryption and signing.  Because SSL and websockets go so great together, I'm finding myself wanting to use openssl for the on-disk crypto as well, rather than shoehorn PGP in there and have to manage two cryptosystems.  I could use PGP for both wire and disk, but I'm concerned about message rates -- perhaps hundreds or thousands of messages per second, with most messages being generated and parsed by machine.  This is not exactly in the sweet spot of PGP's intended use case. If I pretend that PGP never existed in the first place, I'm finding that using openssl is a great way to do wire encryption, counterparty authentication, and message and file signing and encryption.  I'm even thinking I could, just maybe, build an entire web-of-trust-like SSL certificate infrastructure right into my shiny new application, encourage signing parties among users to populate it, and bypass the SSL certificate racket. But that feels so... redundant. below points out the NaCl library as another alternative; has a python library for it: https://github.com/seanlynch/pynacl. What am I otherwise missing, either pro or con?

• Answer:

  OpenSSL is a perfectly acceptable solution to this problem.  PGP is designed for encrypting and signing messages between people.  The relative lack of APIs and the Web of Trust model both reflect that.  It's adaptable for other, more automated tasks as well -- the Debian package repository comes to mind -- but it has far fewer options and isn't as automated as OpenSSL. OpenSSL is, as you said, a Swiss army knife.  You need to be more careful with it, as the many options give you many ways to screw up.  But the flexibility is also very powerful.  Unlike PGP, it's primarily designed to be used programmatically, though it can be used from the command line as well.  It is definitely a better choice for the wire than PGP, and it's not worse for on-disk encryption, just slightly easier to get wrong. PGP's web of trust model is sometimes what you want -- perhaps for peer-to-peer? -- but the OpenSSL PKI is more useful most of the time. You might also consider a special-purpose library like NaCl.  NaCl is blazing fast, but isn't very flexible, it's a pain to build, and it can be kind of tricky to get the design right sometimes.  Some mistakes are harder to make with NaCl, but some are also easier.  I'd probably stick with OpenSSL.