[dccp] other proposed AUTH48 changes for DCCP spec

[Eddie Kohler <kohler at cs.ucla.edu>]

Hi all,

Here are the other technical corrections, clarifications, and two or  
three changes we propose to make in AUTH48 for the DCCP spec.


CORRECTIONS AND CLARIFICATIONS

- Options are explicitly allowed to declare how they behave when  
Mandatory.

- The FGSS feature (Section 6.6.4) claimed that retransmitted Change  
options always have the same data; this is not true.  Correct it.

- (8.5) Explicitly show when packets with short seqnos are dropped  
when they aren't allowed.

- (8.5) Explicitly show when short seqnos are extended.

- (13.2) Elapsed Time should be a LOWER bound, not an upper bound.


TECHNICAL CHANGES

- Allow multiple Init Cookies.  The server can send multiple options,  
all of which the client will echo in the same order.  This avoids the  
253-byte option length limit.

- NDP Count option can take up to 6 bytes of data (since sequence  
numbers are 6 bytes long).

- (7.5.3) We advise Sequence Window be set to the number of packets  
in 5 RTTs.  This conflicts with the definition of "active  
connections" (as receiving packets within 5 RTTs), since it leaves so  
little slack.  Say a connection is speeding up, and sends Sequence  
Window packets in 3 RTTs.  This is not insane.  Now assume all those  
packets are lost.  The endpoints can't get in sync for another 2  
RTTs, until the connection becomes inactive.  For that reason, we  
propose to redefine the "active connections" window to three RTTs  
rather than five, introducing some slack.

- (8.1.2, IANA) Rather than talk about "ASCII" Service Codes, talk  
about Service Codes that follow the guidelines we've laid out for  
"SC:".  (Mark)

- (Appendix A) Add the "ack_runlen" variable, which is required to  
get O(1) acknowledgement processing.


Eddie