Some questions

Hello,

I am currently implementing the asap interface of a registrar server  
and have a few questions:

ASAP keep alive procedures:
Is it necessary to start a timer when a registrar sends an asap  
endpoint keep alive message to wait for the corresponding ack? If  
yes, whats the timeout?
How often should a registrar send an asap endpoint keep alive message?
What should happen if a registrar server receives an asap endpoint  
keep alive ack message, which is maybe not expected?
Which timers exactly need to be reset upon reception of an asap  
endpoint keep alive ack message?

ASAP server announces:
Section 3.6 states that a registrar should send out server announce  
messages every (n+1)*T6-Serverannounce on the multicast channel,  
where n is the number of discovered registrar servers via asap server  
announce messages. Why not look up the internal handlespace registrar  
table?

Some general port stuff:
Are there any rules on which ports a registrar server is reachable  
for asap? Does the server needs to be bound locally to the well known  
enrp or asap port or is any other port good enough as well? If bound  
to the well known asap port, would it be allowed to omit sctp  
transport parameter in the asap announce message so that pool users  
and pool elements can reach the server by using the senders sending  
adress of the asap server announce message?

ENRP server id:
The enrp draft says, that no problems have been identified yet when  
there's a duplicate registrar server id in the handle space. Please  
consider the following situations:
1) There are a few registrar servers and two of these registrars have  
the same server id and one of these two registrars fails.
a)The other server with the same id detects that and initiates a  
takeover with an init takeover message in which sender server id and  
target server id are the same. How should the other registrars react?  
To which registrar should they send the init takeover ack?
b) A registrar which is not one of the two with the same id detects  
the fails and initiates a takeover. Any other server who receives an  
init takeover message would not know which server is to be taken over.

2) There are a few registrar servers and within are two registrar  
servers with the same server id, which is also the highest id in the  
pool. Now, if another registrar server fails these two registrar  
servers would eventually both try to take over the failed peer  
because their ids are equal and one is not less than the other.

Also unique registrar server ids would make live easier for  
implementers(meaning me), by storing registrar servers by id and not  
by id plus assoc information or so. :)

Thanks in advance and best regards

	Frank Volkmer