transport protocols : was Re: Motions before the WG

[Peter Sylvester <Peter.Sylvester@xxxxxxxxxx>]

> hand.  We have some very important documents in front of the group, 

I have a few questions to the implementors of the tcp socket protocol which is
used for cmp and also as a subset in TSP.

- The 2510bis draft discusses in length the problem that it is not
  defined how terminates a connection. 

  This is solved explicitely in the CMP transport draft but this
  discussion has never found its way into TSP. 

  So far, all of the TSA implementations that I have seen, end the
  connection after the reponse thus allowing only one exchange of data.
  
  I have not seen an implementation that implements a polling. To
  a certain degree I would assume that polling is generated when
  one can handle several requests over a connection, otherwise I
  would just wait until the response is available. 

- The cmp transport drafts specifies that messages are send to a server,
  and that a server 'replies'. Does this mean that there is always a
  half duplex behaviour? If so, what would be a good behaviour of 
  if a client sends a length field of 1 gigaoctets? If the protocol
  is half-duplex, it cannot answer with an error before receiving
  the data.