At 08:39 AM 6/30/99 +0800, Barney Dalton wrote:
>
>I treid your test at home - miserabel results!
>
>> import java.net.*;
>> import java.io.*;
>> server = new ServerSocket(2048);
>> socket = server.accept();
>
>dialup dialog came up
>(pressed cancel)
>
One thing you might want to try is to turn off automatic dial-up. You can do
this by going into the Control Panel and clicking on the Internet icon.
A multipane dialog appears. Select the Connection pane and then select
the Connect to Internet using a local area network. Once you have done this,
opening a Web browser or pinging no longer triggers an attempt to make a
dial-up connection. You can still make such a connection by clicking on
the dial-up icon that represents your service provider.
>>
>> The bsh should appear to hang as the Java server (i.e., ultimately jdebug)
>> waits for a connection. Next, open a second window on the Emacs *scratch*
>> buffer and execute
>>
>> (open-network-stream "jdebug" "*scratch*" "127.0.0.1" 2048)
>
>dialup dialog came up again
>(pressed cancel, the server.accept() call finished)
>
>I'm using windows95 with a lmhosts with a 127.0.0.1 localhost entry. About
>the only piece of networking that doesn't seem to prompt for a dialup is
>ping 127.0.0.1 even ping localhost seems to require a dialup.
>
>I had a bit of hunt for an answer most people seem to refer to the
>hosts/lmhosts files, but I think we've established that there is a bit
>more to the problem. There does seem to be some mention of upgrading to
>winsock 2 to solve the problem - I think win98 has this, does it suffer
>the same problem?
>
Despite all the claims to the contrary, I believe Winsock no longer uses
hosts or lmhosts, if it ever did, at least on Windows 95. My only
experience with Windows/NT is at work and of course at work we're
networked, so this is never an issue.
>
>My 2c worth on the socket interface.
>
>i) Does the use of sockets represent a departure from normal emacs coding?
>If yes then can't a more commmon technique be used. I suspsect that the
>answer is no and that sockets have been used a fair bit.
>
Standard I/O is the traditional channel of communications between
Emacs and external processes, such as debuggers. The only application I know
of that uses a socket connection is browse-url, for obvious reasons.
>ii) Would having to use something other than sockets limit the scope of
>the project? If so I think it would be a great pity to limit it purely
>because microsoft didn't have time to implement a proper TCP/IP stack. I'm
>sure a work around will be found, and emacs users aren't averse to the odd
>installation challenge (even if they do use win95!).
>
Using a socket connection simplifies implementation. The only thing it would
allow that couldn't be done with standard I/O, AFAIK, would be for Emacs to
talk to an instance of the debugger running on another machine.
My inclination at this point is to go with a socket because it avoids the need
to mutiplex debugger and application traffic across the same channel.
- Paul
