🚀 go-pugleaf

Vladimir Stefanovic wrote:

> 1) Should I send data immediatelly, while I'm still in
> ClientExecute()?

Yes.

> 2) If yes, how can I address the client? Maybe the 'Stream->'
> pointer (I think that Mr Hans Galema said that once to me)?

Indeed I told you already how to do that.

<old reply>
Date: Mon, 27 Sep 2004 21:53:51 +0200
From: Hans Galema <dontusethis@dontusethis.nl>
Newsgroups: borland.public.cppbuilder.internet.socket
Subject: Re: TClientSocket/TServerSocket data exchange...
References: <41585538@newsgroups.borland.com>
In-Reply-To: <41585538@newsgroups.borland.com>

Vladimir Stefanovic wrote:

 > I want my Client to send a request to Server to get
 > some file. Imaginary protocol:
 >
 >    GET filename.txt

Allow rigth for binary files too:

	GET filename.pdf

 > Then, I want the Server to receive the request and to
 > send back the Client the desired file contents. Imaginary
 > protocol:
 >
 >    FIL this_is_the_pure_ascii_contents_of_the_filename.txt

Better:
	size FIL this_is_the_pure_contents_of_filename.doc

or

	FIL size this_is_the_pure_contents_of_the_filename.doc

The last will do if the client only expects 3 char commands.

 > The Server receives data through it's own thread
 > stThreadBlocking & ServerSocketGetThread().
 >
 > How I'm supposed to SEND BACK the Client the desired
 > file? Where I have to place the sending code? Should I
 > have the alternative TClientSocket on Server to use for
 > sending? There are some examples of how to send a row
 > of text back ( SendText() ) but my file can be several KB
 > so I cannot send the file back from only one pass...

Nothing of that all.

You got data from a client using something like
	
	pStream->Read()

in your Threads Execute() method. In the same method you can use

	pStream->Write()

to send data to the client.
</old reply>

> 3) Should I use the same concept of sening data like I did
> from the client side (piece by piece)?

Mostly yes.

> 4) On the clients side I suppose that I should use on OnRead(),
> what else? And here comes the main of my concerns:

If it is non-blocking yes. Sending/Receiving with Client/Server
is the same. There is only difference with blocking/non-blocking.

> 5) AFAIK, clients OnRead() is the common passage for all
> reads (it's not thread), so I have to be able to:
>    - start receiving
>    - extract 'sizeof_stream' when the received ammount of data
>    exceeds sizeof(int)
>    - continue reading while TOTAL_RECEIVED == sizeof_stream,
>    then prepare for new reads
>    - Clear-up if some error occure and prepare for new read
> 
> Does anyone have some similar code I should look at?

Such code or atleast part of such code has often been posted here.

> Sometimes I think that it's easier to have alternative
> TClientSocket on server for sending and TServerSocket
> on client for receiving...

No. Keep on. You will do it the normal way.


>    InboxDataBuffer = new TMemoryStream;

For readibility: do not call that a databuffer.
It is an instance of s TMemoryStream, so better:

      InboxMemoryStream = new TMemoryStream;



>       if ( NewSize == expected_size )

That will not do. The server can send several messages.
If the server sends two messages:
The client receives just pieces. Once piece can consist
of the last bytes of the first message and the first bytes
of the last message.

So for determining the size ? It is not at byte 0 of your stream
for the second message now.

Better:

	if ( NewSize >= expected_size )


But yet better. Fill the memorystream just for one message,
handle every messaga separately, and then continue reading with
a clered stream.	

Hans.