[jcifs] jCifs writing to Mailslots

[Sascha Teifke]

Christopher R. Hertel wrote:

>On Tue, Jun 10, 2003 at 04:42:54PM +0200, Sascha Teifke wrote:
>  
>
>>Yes, they do.
>>    
>>
>
>Only if there is something listening.  Even on older Windows systems, you 
>need to have the WinPopup application running or the messages are not 
>displayed.
>
My world is a 100% pure WinXP Environment :-( I'm not very happy about 
this but it is as it is ...
To receive those messages you need to run a Service called Messaging 
Service (translated from german)
who runs by default and isn't deactivated in "my world".

>There is a newer RPC-based mechanism that is more reliable, but I don't
>know much about it.
>  
>
That sounds very interesting. I will have a look on it!

>There was some company (in Florida, I think) that was selling software
>that would send on-screen SPAM-a-grams using the RPC technique.  The
>software would connect to port 135 (which is the DCE/RPC endpoint
>resolution service), and then perform an MS-RPC call to cause the popup to
>happen.  Most ISPs were blocking the "normal" NBT and CIFS ports, but not 
>135, so the SPAM-a-grammers were able to pop advertisements onto peoples 
>Windows desktops.
>
>Ick.
>
>
>  
>
>>IMHO, Netsend Messages are useful if you need to alert an amount of 
>>Win32 users without installing any client software
>>who needs to poll a specific network server.
>>    
>>
>
>Only if they're running the WinPopup application so that they can receive 
>the messages.  Also, it doesn't do much for MacOS, PalmOS, or *nix users.
>
>  
>
>>One Problem is, that there is no way to send messages to a specific list of
>>users at once.
>>    
>>
>
>Well, not unless you can retrieve a list of users in a group.  If you can
>get hold of the list then you just iterate through it... but you knew
>that.
>
>  
>
>>If you try to make a script that executes every single 
>>"net send" to shell you often have to wait a long time
>>because if a computer is turned off it may last up to a minute until it 
>>returns a message. Parallel processing of "net send"s
>>has made me mad - it seems like it isn't thread safe.
>>    
>>
>
>You could do this on a *nix box without much trouble using smbclient and 
>nmblookup.  One problem with the smbclient utility is that it doesn't 
>look up the name correctly (at least, that's true for the version I'm 
>testing with), so you have to do the name lookup using nmblookup first.
>
>How have you tried scripting this?
>
>  
>
>>The environment I'm working in doesn't allow me to write any kind of 
>>client Software that could be installed on every client.
>>So I need to use the standard unrelieable way.
>>    
>>
>
>Sounds typical.  What a pain.
>
>  
>
>>Sascha
>>
>>Allen, Michael B (RSCH) wrote:
>>
>>    
>>
>>>Mmm, yeah. Unfortunately it's just API documentation which isn't that 
>>>useful at this point.
>>>      
>>>
>
>Point.
>
>  
>
>>>Looking at a quick capture it appears as though there is no session 
>>>setup or tree
>>>connect at all. It's just an NBT session establishment message followed by 
>>>a minimal
>>>SMB the format of the actual payload is as trivial as described below.
>>>      
>>>
>
>Mailslot protocol is like this as well.  It uses port UDP/138 but formats 
>messages in SMB format.  The header is almost all empty, as I recall.  
>There's no NBT session request because it's UDP, not TCP.
>
>  
>
>>>This would require
>>>modifying the jcifs.smb files although it would not be difficult. The 
>>>question is, why
>>>would one want to use this? It's a neat trick but it's not reliable. Does 
>>>W2K or XP still
>>>support these messages?
>>>      
>>>
>
>Backward compatibility with older Windows systems is generally, but not 
>uniformly, supported.  This stuff is so touchy...
>
>Chris -)-----
>  
>