[evla-sw-discuss] VCI and Message IDs

[Martin Pokorny]

Bryan Butler wrote:
 > the decision on whether to use UDP or TCP (or other) in the
 > communications between the rest of the system and WIDAR has not been
 > made, that i'm aware of.

 From my perspective as a rank VCI novice, it seems that the issue of 
UDP vs TCP is sort of beside the point here. From what I've read, it 
seems that there is one aspect of the VCI design that determined the 
request/response style of communication and the inclusion of message 
IDs: asynchronous VCI "calls". The need to support an asynchronous call 
style is apparent from the piecemeal nature of creating a complete 
configuration; that is, a configuration can consist of more than one 
configuration message. Not until the "mapping time" occurs, can the 
configuration request be truly validated. I don't think that you would 
want the requesters to block until the mapping time occurs. Requesters 
do get a simple ACK/NACK back immediately after sending the request, and 
this seem right, but that's not the end of the configuration process.

Sonja also writes that the two communication endpoints are peers, not in 
a client/server relationship, which is appropriate when handling 
asynchronous calls. That might argue against using HTTP, but that's a 
somewhat different issue.

-- 
Martin