[dccp] Re:

[Eddie Kohler <kohler at CS.UCLA.EDU>]

Hi Somsak,

On Jul 11, 2005, at 12:42 AM, Vanit-Anunchai, Somsak - VANSY014 wrote:

> Hi
>
> I've some questions regarding draft-dccp-spec-11.ps. Hope you can  
> give me some advice.
> 1. Page 53 above item 8.31
> “DCCP-Data,DCCP-DataAck and DCCP-Ack packets received in CLOSEREG  
> or CLOSE states MAY be either processed or ignored.”
> CLOSE should be CLOSING right ????

Yes, you're right!  Good eye, thanks.

> 2. The application data can be sent out during connection  
> establishment with DCCP-Request or DCCP-Response packet.
> Could that application data be either forwarded immediately to the  
> user or stored in an buffer, after connection established then  
> forwarded to the user?
> I suspect it could be both.
> Could you please confirm?

This is correct, although I would think most implementations would  
buffer it.

> 3. Page 24 item5.5
> “As with DCCP-Ack packets, DCCP-CloseReq and DCCP-Close packets MAY  
> have non-zero-length application data areas, whose contents  
> receivers MUST ignore.”
> If receivers must ignore the application data, why do we send it out??

We don't require that you put anything in the application data  
area!!  But packets with non-empty data areas are not erroneous.   
This is for consistency, and for uses such as path MTU detection.

> 4. From the data transfer perspective, it seems to be the  
> connectionless UDP. If so, what is the purpose of connection  
> establishment and teardown?

The three-way handshake "guarantees" the presence of both parties,  
which makes some DoS attacks harder to execute.  There are other  
reasons, including firewall traversal.

Thanks,
Eddie


> Many thanks
> Somsak