time stamping interoperability tests

[Peter Sylvester <Peter.Sylvester@xxxxxxxxxx>]

Hello,

during the last two months I had a chnce to make a few
interoperability tests with some implementations of
experimental services. There haven't been many so far.

I would like to share the few results here so in order
to allow other implementors to gain some time.

Socket protocol: 

- The first 4 octets need to be in network byte order.   

- An errMsg response (X'06') is not a possible response to
  a request according to the protocol specification.

- The connection is immediately closed after having sent the
  response. This is a possible interpretation of the protocol, 
  but seems unfortunate to me, since it doesn't make it possible
  to make several requests through the same tcp connection.
  Note that during CMP interoperability testing a similar
  problem occured, an internet draft describes this.

- Implementations did not implement any time out for the TCP
  connection. 

TSP encoding :

- Instead of encoding a bit string, one implementation encoded
  the position of one error bit as an integer.

- One implementation did not accept a NULL parameter for the digest algorithm. 

- So far, I haven't send really nasty requests with valid or invalid
  arbitrary asn1 structures, but:

  One implementation does not parse the request correctly,
  a symptom was that the reqCert flag is not detected. 
  The actual case was that due to a change in the syntax 
  between some drafts that I missed the request did not encode
  correctly the tsaPolicy, instead of accepting the request, 
  an encoding error should have been signaled I guess.

  Note, that the format of the request and response have
  changed in incompatible ways several times. 

Peter Sylvester