On Mon, 2008-04-07 at 12:52 -0400, Joshua Kramer wrote: > > I will be making a (hopefully) small number of changes to the existing > > network IO code infrastructure in the next couple of weeks which should > > help this happen. > > After reviewing the NSS documentation and considering this, I think it's > best to create a completely separate SSL class that is somewhat like the > normal TCP/IP class. As far as I know, a socket must either be SSL or not > from the time it starts listening - i.e. you can't have a normal socket > switch to SSL if it detects an encrypted connect attempt. > > Is there anything I should consider about your new model as noted above, > or can I model the SSL class on existing code?
I think you're on the correct track. I don't think that what I'm doing should need too much consideration from you. If you like I'll happily look over code/a patch and let you know what I think. Andrew
