Prompts the SysOp for system password.
If a valid username is sent and IP is not in rlogin.can and the local username
wasn't the correct password, prompt for password.

requests to allow differentiation between received responses (ACK/NAK) and
remote option requests. This solves the problem with Kermit-95 waiting forever
for a response (ACK or NAK) to its initial option requests (i.e. "Negotiate
About Window Size" and "New Environment Option").

1) RLogin connections are accepted from any IP
2) If the IP is not in the rlogin.can, the 2nd user name is tested against the
   users password.  If it matches, the user is auto-logged on.
3) Currently, the case where the name exists but the password doesn't match
   isn't what I'd like it to be... ideally, it should go straight to the password
   prompt.  It currently assums a new user... mostly because I wanted to keep
   this change easy to back out of anyone hates me.

* handles long (>25 char) user names better
* log message when logging in as an unknown user
* log message when no user name specified during login
* no longer send telnet negotiation (assuming Telnet session) if RLogin fails

Sets the terminal variable to "ansi", "rip", "wip", "html", or "dumb" if it's
not otherwise initialized via Telnet or RLogin.

(pth) since incom() blocks here.

old 486 hardware!

of rows for Windows 2K/XP telnet client (reports 105 lines without the char).

variants (gmtime_r, localtime_r, and ctime_r) for thread-safe operation on Unix

Win2K/XP telnet).
Corsor position report logged.

This increases the likelyhood of detecting RIPscrip support properly in RIPtel.
Also changed the extra response info delay method (to 1 yield per char rx'd).

Moved the telnet negotiation sequence BEFORE the terminal-detection sequence (fixes the auto-ANSI detection with CRT).

Each function that requires a temp string variable, now has its own "tmp" local variable.