Geoff wrote:
Or we write our own BIO that wraps up whatever glue is desirable on the wine-side. Worst-case (and it's not that bad an alternative) is that you use a memory-base BIO to encapsulate all I/O and worry about moving data to and from "sockets", whatever they look like, afterwards.
That's what I've always done. And it seems I've managed (via the rant at http://kegel.com/ssl/api.html) to convince Sun to make that the standard way to do ssl with NIO under the next JDK, kinda. So it's not a bad way to go at all. - Dan