I don't believe its a non-sequitur because its architecturally addressing some of the space of the problem and intrudes into the API without requiring a name. The implication all activity in the path between communicating entities require names to be used seems to me to be somewhat moot. I used it as an example to discuss two specific methods of dealing with the question of how different application spaces cope with these situations. Shims and wraps are common.
The question(s) were posed 'how would ssh..' or 'how would ftp...' and I replied, contextually how I feel other people have approached the problem. What I took from their approach is: they found a way to do it without magic labels in the DNS
I could have mentioned tun/tap too.
-G
On Tue, Jul 21, 2015 at 1:26 PM, Ted Lemon <ted.lemon@xxxxxxxxxxx> wrote:
On Jul 21, 2015, at 1:36 AM, George Michaelson <ggm@xxxxxxxxxxxx> wrote:SOCKS works with a shim. there is no .SOCKS domain to make SOCKS work.This is a non-sequitur. SOCKS is a tunnel for your network API. Architecturally, it’s quite a poor choice for solving the problem we are discussing. It’s used to solve that problem because it’s the easiest hack to make it work, not because it’s the right thing to do.