On Tuesday 24 February 2015 14:48:34 Jon A. Cruz wrote: > Another thing to consider is the route that OpenSSL has taken. They have > exposed a few calls to register mutex/thread handling, but don't use any > internal implementation themselves. The hosting app is then free to > initialize and configure things as appropriate.
That's what libdbus-1 did and we had to redesign. That kind of interface is horrible for a library that gets used by other libraries. I advise against it. > IoTivity could expose a simple thread support API like OpenSSL does, and > include a default implementation if a system/application does not need > anything custom. > > Or... we could at the least ensure that our internal wrappers are used > in a way such that a public API for locking can be exposed. That sounds > like a helpful separation. -- Thiago Macieira - thiago.macieira (AT) intel.com Software Architect - Intel Open Source Technology Center
