Telnet: Grabbing Bytes Previously Written During Next Read?

[pbd22]

Hi.

I am building a custom telnet interface and my problem is that I
want to read the user input along with the previously written stream.

Right now I am logging the user.

I have

Login: Bill

Login is written by:

buffer = ASCII.GetBytes("Login: ");
_clientStream.Write(buffer, 0, buffer.Length);

Bill is entered by the user.

At the next return, I want to read:

"Login: Bill".

This is how I know its a Login read
and not a Password: read.

I am reading input with the following loop:

Code:
while (true)
{

_bytesRead = 0;
try
{
//blocks until a client sends a message
_bytesRead = _clientStream.Read(message,
0,4096); //message

}
catch
{
//a socket error has occured
break;
}

if (_bytesRead == 0)
{
break;
}

statusMessage += ASCII.GetString(message, 0,
_bytesRead);
But, my problem is that I get:

"Bill".

when the user submits his string.

How do I get both the written "Login: " and
the read "Bill" on the same read?

Please let me know if i need to provide more information,
otherwsie, thanks a lot for your response!

[Peter Duniho]

On Wed, 16 Jan 2008 09:57:27 -0800, pbd22 <du*****@gmail.comwrote:

[...]
At the next return, I want to read:

"Login: Bill".

This is how I know its a Login read
and not a Password: read.

[...]
But, my problem is that I get:

"Bill".

when the user submits his string.

How do I get both the written "Login: " and
the read "Bill" on the same read?

Maybe it's been too long and I'm misremembering how the telnet protocol
works, but my recollection is that the telnet client is not supposed to
echo the data you send it.

In other words, there will never be a "Login: " for your telnet server to
read. You need to track the state of the connection explicitly, by
knowing what you've sent to the client, as it relates to what input you've
received back.

For example: assuming you are sure you've processed all user input up to
this point (for a login, this should be trivial :) ), then when you senda
login prompt, you need to set some local state so that you know the next
thing you should receive from the user is the login ID.

Once you have successfully processed the user's login ID response, then
you would send the password prompt and set some local state so that you
know the next thing you should receive from the user is a login password..

The "state" could as simple as an enum that describes the various stages
of a client connection:

enum ClientState
{
LoginIDPrompt,
LoginPasswordPrompt,
Connected
}

and then a switch() statement where you process user input that uses a
variable associated with the client assigned to a value from the enum to
decide what to do with the input:

// at this point, having read an entire line of client
// input and stored it in a variable (say, "strUserInput"
// for the sake of discussion):
switch (clientData.ClientState)
{
case ClientState.LoginIDPrompt:
strLoginID = strUserInput;
clientData.ClientState = ClientState.LoginPasswordPrompt;
break;
case ClientState.LoginPasswordPrompt;
if (CheckPassword(strLoginID, strUserInput))
{
clientData.ClientState = ClientState.Connected;
}
else
{
SendLoginPrompt();
clientData.ClientState = ClientState.LoginIDPrompt;
}
break;
case ClientState.Connected:
// handle regular user input here
break;
}

That's just the basic idea. There are lots of ways to actually implement
it.

Pete

[pbd22]

Thanks Pete.

I am making my way though your code.
Could you explain what the "clientData" of "clientData.ClientState"
is?

Sorry - not too familiar with enums.

Thanks,
Peter

[pbd22]

Thanks Pete.

I am making my way though your code.
Could you explain what the "clientData" of "clientData.ClientState"
is?

Sorry - not too familiar with enums.

Thanks,
Peter

[pbd22]

Thanks Pete.

I am making my way though your code.
Could you explain what the "clientData" of "clientData.ClientState"
is?

Sorry - not too familiar with enums.

Thanks,
Peter

[pbd22]

Thanks Pete.

I am making my way though your code.
Could you explain what the "clientData" of "clientData.ClientState"
is?

Sorry - not too familiar with enums.

Thanks,
Peter

[Peter Duniho]

On Wed, 16 Jan 2008 10:43:33 -0800, pbd22 <du*****@gmail.comwrote:

Thanks Pete.

I am making my way though your code.
Could you explain what the "clientData" of "clientData.ClientState"
is?

"clientData" is just a variable name I made up. You should have in your
own code some specific variable that contains state information about each
client connection (such as the _clientStream variable). You may in fact
put code similar to what I posted _inside_ the class that contains this
state (the fact that the _clientStream variable is unqualified suggests
you may already be doing this), in which case you wouldn't need the
"clientData" part at all. You'd just refer to the "ClientState" member
directly (or whatever you choose to call it).

Sorry - not too familiar with enums.

I apologize for any confusion. The "clientData" doesn't really have
anything to do with the enum aspect. The "ClientState" is simply a member
of a data structure (probably class, but could be a struct instead) that
has the type ClientState, and "clientData" would simply be a reference to
an instance of that data structure.

For that matter, don't get too hung up on the enum. I used an enum
because it's a convenient, readable way to describe simple state like
this, but you can manage the state however you like.

Pete

[pbd22]

Yes, I think I am already handling state information at the head of my
TheConnectionHandler
with the following lines of code:

_telnetSocket = (TcpClient)connectionQueue.Dequeue();
_clientStream = _telnetSocket.GetStream();
Here is the core of the handler and the Login method per your
suggestion. Does this look right?

Thanks again.

public void TheConnectionHandler()
{

_telnetSocket = (TcpClient)connectionQueue.Dequeue();
_clientStream = _telnetSocket.GetStream();

//if (IsLoggedIn == false)
//{
// buffer = ASCII.GetBytes("Login: ");
// _clientStream.Write(buffer, 0, buffer.Length);
//}

while (true)
{

_bytesRead = 0;
try
{
//blocks until a client sends a message
_bytesRead = _clientStream.Read(message,
0,4096); //message

}
catch
{
//a socket error has occured
break;
}

if (_bytesRead == 0)
{
break;
}

statusMessage += ASCII.GetString(message, 0,
_bytesRead);

if (statusMessage.LastIndexOf(ENDOFLINE) 0)
{

buffer = ASCII.GetBytes(Login(statusMessage));

}
}

_telnetSocket.Close();

}

protected string Login(string clientInput)
{

switch (ClientState)
{
case ClientState.LoginIDPrompt:
strLoginID = clientInput;
clientData.ClientState =
ClientState.LoginPasswordPrompt;
statusMessage = "";
break;
case ClientState.LoginPasswordPrompt:
if (CheckPassword(strLoginID, clientInput))
{
clientData.ClientState =
ClientState.Connected;
}
else
{
SendLoginPrompt();
clientData.ClientState =
ClientState.LoginIDPrompt;
}
break;
case ClientState.Connected:
// handle regular user input here
break;
}

[Peter Duniho]

On Wed, 16 Jan 2008 11:36:52 -0800, pbd22 <du*****@gmail.comwrote:

Yes, I think I am already handling state information at the head of my
TheConnectionHandler
with the following lines of code:

_telnetSocket = (TcpClient)connectionQueue.Dequeue();
_clientStream = _telnetSocket.GetStream();
Here is the core of the handler and the Login method per your
suggestion. Does this look right?

Mostly. Some suggestions:

* Don't have the Login() method manage the "statusMessage" variable..
Your "TheConnectionHandler()" method is a more appropriate place to do
that, as it's what is manipulating it otherwise.

* You should use the index of the EOL for your user input to createa
substring that's passed to the Login() method, to ensure that that method
gets _only_ the one line of user input. TCP being what it is, it's
possible to receive additional data after the EOL as part of the same
chunk of read data. In the user login scenario hopefully the chances are
less (presuambly it'd require the client to type ahead, entering both the
login ID and the password all at once, before any of that data actually
gets sent...not impossible, but probably unlikely). But you shouldn't
leave something like that to chance.

* And combining the two above, obviously this means that you shouldn't
just clear the "statusMessage" variable to an empty string. You should
extract the data you're ready to process, and reassign the remaining data
to the variable (so basically two calls to Substring() to split up the
string).

Of course, you left in the "clientData" variable from my original
suggestion...if you're just storing a ClientState field or property, then
anywhere you've written "clientData.ClientState", you probably just want
"ClientState". I assume that's just an oversight that would be addressed
in the actual code (since it wouldn't compile otherwise).

Pete

[pbd22]

Thanks Pete.

I feel a bit thick on this.

I am still struggling with ClientSate in the switch statement.

Left alone, I get

"ClientState is a type but is used like a variable".

When I do something like

ClientSate state = new ClientState();

or

state = new ClientState();

and then

switch (ClientState)

the compiler still throws the error.

shouldn't i be instantiating the connection
as a new ClientSate? Or something like that?

What am I not getting?

ps -

I have put:

enum ClientState
{
LoginIDPrompt,
LoginPasswordPrompt,
Connected
}

inside my class next to my
variable declarations.

[Peter Duniho]

On Wed, 16 Jan 2008 12:25:27 -0800, pbd22 <du*****@gmail.comwrote:

Thanks Pete.

I feel a bit thick on this.

That's okay. I know how hard it can be to try to write code on a Monday..
:)

I am still struggling with ClientSate in the switch statement.

Left alone, I get

"ClientState is a type but is used like a variable".

That's because you (apparently) do not have a variable named "ClientState".

When I do something like

ClientSate state = new ClientState();

Variable name is "state", not "ClientState".

or

state = new ClientState();

See above.

and then

switch (ClientState)

the compiler still throws the error.

Sure, it would. If you've named your variable "state" instead of
"ClientState", then the switch statement should read "switch (state)"
instead of "switch (ClientState)".

Again, this isn't unique to enums...don't get hung up the fact that it's
an enum. The point of using an enum is that they can be used like a
regular simple value type. So the rules for writing code using enums are
the same as if you were writing code using, for example, an int or a char
or a long, etc.

I'm sorry if using the same name for the variable as the type has confused
things. It's funny...I generally try _not_ to do that, but it's a common
..NET convention for naming properties and public fields. So when I had a
data member of a data structure of type "ClientState" in my original
example, I just reused the type name for the name of that data member.
Things seem to have gone downhill from there. :)

Pete

[pbd22]

Hey Peter,

OK, I am starting to get the swing of things, thanks.
The last bit I am having probs with is the 2 substrings.
My current program is going screwy once it connects.
It either hangs, or says your connected after I type each
letter in the Password: field.

I think it has to do with the statusMessage variable.
I think this last bit will do it.

Thanks again.
Peter
public void TheConnectionHandler()
{

_telnetSocket = (TcpClient)connectionQueue.Dequeue();
_clientStream = _telnetSocket.GetStream();

buffer = ASCII.GetBytes("Login: ");
_clientStream.Write(buffer, 0, buffer.Length);

while (true)
{

_bytesRead = 0;
try
{
//blocks until a client sends a message
_bytesRead = _clientStream.Read(message,
0,4096); //message

}
catch
{
//a socket error has occured
break;
}

if (_bytesRead == 0)
{
break;
}

statusMessage += ASCII.GetString(message, 0,
_bytesRead);

if (statusMessage.LastIndexOf(ENDOFLINE) 0)
{

if (currentState != ClientState.Connected)
{
Login(statusMessage.Substring(0,
statusMessage.IndexOf(ENDOFLINE)));
//buffer = ASCII.GetBytes(statusMessage);
}
else
{
buffer = ASCII.GetBytes("You are connected!");
_clientStream.Write(buffer, 0, buffer.Length);
_clientStream.Flush();

}
}

}

_telnetSocket.Close();

}

protected void Login(string clientInput)
{

switch (currentState)
{
case ClientState.LoginIDPrompt:
strLoginID = clientInput;
currentState = ClientState.LoginPasswordPrompt;
statusMessage =
statusMessage.Remove(statusMessage.IndexOf(clientI nput),
statusMessage.LastIndexOf(clientInput)); < --- SHOULD I DO THIS HERE?
//statusMessage = "";
SendPrompt();
break;
case ClientState.LoginPasswordPrompt:
if (CheckPassword(strLoginID, clientInput))
{
currentState = ClientState.Connected;
}
else
{
currentState = ClientState.LoginIDPrompt;
}
break;
case ClientState.Connected:
{
break;
}
}

}

private void SendPrompt()
{
buffer = ASCII.GetBytes("Password: ");
_clientStream.Write(buffer, 0, buffer.Length);
_clientStream.Flush();

}

public static bool CheckPassword(string name, string pass)
{

if (name == "First" && pass == "Last")
{
return true; // just for kicks. remove it.
}
else
{
return false;
}

}

[Peter Duniho]

On Wed, 16 Jan 2008 14:04:23 -0800, pbd22 <du*****@gmail.comwrote:

Hey Peter,

OK, I am starting to get the swing of things, thanks.
The last bit I am having probs with is the 2 substrings.
My current program is going screwy once it connects.
It either hangs, or says your connected after I type each
letter in the Password: field.

I think it has to do with the statusMessage variable.
I think this last bit will do it.

Well, this is related to my suggestion that you only manage the
statusMessage variable from within the method "TheConnectionHandler()".
The code you put in the Login() method to deal with updating the
statusMessage variable is wrong in any case (IndexOf() and LastIndexOf()
in that statement are both returning the same index...nothing's really
getting removed from the string), but IMHO the code will be easier to
understand and fix if you put that statement in the method dealing with
i/o, rather than the one dealing with login credentials.

IMHO, you should remove the line that attempts to deal with updating the
statusMessage variable from the Login() method. Then the section of code
in TheConnectionHandler() that deals with the data after it's been
converted to a string should look more like this:

// You want the first EOL in the string, not the last.
int ichEOL = statusMessage.IndexOf(ENDOFLINE);

// If there is an EOL character, then process the current line
if (ichEOL >= 0)
{
// Extract the line of text, up to the EOL (but not including)
string strLine = statusMessage.Substring(0, ichEOL++);

// If there are characters after the EOL, set the current
// string to those characters. Otherwise, just reset it to
// an empty string.
if (ichEOL < statusMessage.Length)
{
statusMessage = statusMessage.Substring(ichEOL);
}
else
{
statusMessage = "";
}

if (currentState != ClientState.Connected)
{
Login(strLine);
}
else
{
buffer = ASCII.GetBytes("You are connected!");
_clientStream.Write(buffer, 0, buffer.Length);
_clientStream.Flush();
}
}

Now, all that said, I'll note that you haven't really got a genuine telnet
implementation there. As you saw when dealing with the entering of the
password, data is sent as the user types it. It's not being sent a line
at a time, rather it's being sent a character at a time.

As long as the data is entered without errors, the above should work
fine. But as soon as you have a situation where the user wants to
backspace, you've got a problem. The backspace characters are just going
to get added to the input string, and so of course when the user finally
does send the EOL character, the string up to that point will be messed up.

IMHO, you would be better off refactoring the code so that you've got two
layers: a low-level telnet i/o layer that can deal with backspaces and the
line, and a higher level layer that deals only in complete lines. The
lower level wouldn't present a complete line to the higher level until it
sees an EOL character.

Whether you split the code up like that or not, you definitely need to
include some logic to deal with the user entering backspaces (and possibly
other control characters as well).

Pete

[pbd22]

Pete -

Thanks,

I agree with you about the "character-by-character" implementation.
When I enter my password things get screwy. I need to figure this
out.

Your code sort of worked. I get all sorts of garbage when I get to the
password ("/n") or ("/r") or ("r/n/r"). These are usually before the
actual
password. I think it may have to do with the fact that the ENDOFLINE
variable (that you can't see) is actually

const string ENDOFLINE = "\r\n";

Maybe this is messing things up?

Peter

[pbd22]

Pete -

Thanks,

I agree with you about the "character-by-character" implementation.
When I enter my password things get screwy. I need to figure this
out.

Your code sort of worked. I get all sorts of garbage when I get to the
password ("/n") or ("/r") or ("r/n/r"). These are usually before the
actual
password. I think it may have to do with the fact that the ENDOFLINE
variable (that you can't see) is actually

const string ENDOFLINE = "\r\n";

Maybe this is messing things up?

Peter

[pbd22]

OK,

You are right. I need to figure out how to get the Password line to
react to lines and not characters...

[pbd22]

OK,

You are right. I need to figure out how to get the Password line to
react to lines and not characters...

[Peter Duniho]

On Wed, 16 Jan 2008 15:16:01 -0800, pbd22 <du*****@gmail.comwrote:

[...]
I think it may have to do with the fact that the ENDOFLINE
variable (that you can't see) is actually

const string ENDOFLINE = "\r\n";

Maybe this is messing things up?

Sure, that's definitely going to be part of it. I didn't realize that was
a string instead of a character and the code I posted assumes it's a
single character. That's easy enough to change though. Just change the
"ichEOL + 1" to "ichEOL + EOL.Length".

As you've noted, there are other things you'll need to address as well.
But that should at least fix that issue. :)

Pete

[pbd22]

Thanks Pete.

I am getting there thanks to your help.

I have some nick-knacks that are, it seems,
not that trivial.

1) how do I highlight a line?
2) how do I change background color?
3) how do I change font color?
4) how do I clear the screen (aka. DOS "cls");

For anybody who cares to tackle 'em...

Thanks again for your help,
Peter

[pbd22]

One more to the list for anybody who might know...

1) how do I highlight a line?
2) how do I change background color?
3) how do I change font color?
4) how do I clear the screen (aka. DOS "cls");
5) how to hide password entry (as clear-text or XXXXXXX)?

Thanks again,
Peter

[Peter Duniho]

On Thu, 17 Jan 2008 09:01:31 -0800, pbd22 <du*****@gmail.comwrote:

Thanks Pete.

I am getting there thanks to your help.

Glad it's working out. You're welcome.

I have some nick-knacks that are, it seems,
not that trivial.

1) how do I highlight a line?
2) how do I change background color?
3) how do I change font color?
4) how do I clear the screen (aka. DOS "cls");

These are all new questions, and deserve an all-new thread.

Hint: when you repost those questions, you should be clear about how you
are displaying the text. There's a huge difference between doing those
kinds of things in a console window and doing them in a normal Forms
application window. For the former, see the Console class (it has
properties for setting font foreground and background, and a method for
clearing). For the latter, you can use a RichTextBox as a quick-and-dirty
implementation, or you can handle all of the drawing yoursself (more
complicated, but gives you more control and in the context of something
like this where the text is constantly changing, generally will provide
better visual quality and performance).

It may be you can get started with the above information. If you need
more help, start a new thread with specific questions.

Pete

[pbd22]

OK, good point. I'll start another thread.

Thanks a ton Pete for your help!

Peter

[christery]

4) how do I clear the screen (aka. DOS "cls");

Aha, another one missing chr$(12)... in this case its some <esc>[2J or
something... who knows, who cares
what do I know... anyone seen chr$(12)
//CY