Network Working Group                                        R. T. Braden
Request for Comments #238                                    UCLA-CCN
NIC #7663                                                    September 29, 1971
Category:
Updates: RFC #171, RFC #172
                   COMMENTS ON DTP AND FTP PROPOSALS
   Data Transfer Protocol
   ----------------------
   1. In the Descriptor/Count mode, the Information Separators should
   have a transaction sequence number field.  Otherwise, the receiver
   cannot be sure he received all transactions before the separation.
   This requires that there be two forms of information separators, one
   for Descriptor/Count mode, the other for the DLE mode.
   2. The modes-available handshake should not be mandatory, as it makes
   no sense in the simplex case.  The receiver doesn't care what modes
   the transmitter _might_ use; he only cares what mode _is_ used, which
   he discovers when the first data or control transaction arrives.  Even
   in the duplex case, it is not clear what use the receiver should make
   of the modes-available information from the transmitter.
   File Transfer Protocol
   ----------------------
   1. The protocol allows an end-of-file to be indicated by closing the
   connection.  This is the same mistake which we made in an early
   version of NETRJS.  Closing the connection without a File Separator
   transaction should only be used to indicate an error, i.e., to abort
   the transmission; it should never be used to indicate normal
   completion of file transfer.  The reason is obvious: there is no way
   for the receiver to tell whether CLS indicates normal completion or an
   abnormal condition in the other host (e.g. the file transfer program
   died).
   2. There should be two forms of the _store_ request, one which fails
   if a file of the same name already exists, and one which replaces an
   existing file of the same name (as now).
   3. A service center host may be expected to require username and
   password transactions before any others are accepted.
   4. There are no error transactions defined for lost data or lost
   synch.  It is assumed there are handled at the DTP level?
   5. All of the defined error codes should be allowed (and encouraged)
   to have explanatory text following them.
                                                                [Page 1]

RTB:gjm
       [ This RFC was put into machine readable form for entry ]
       [ into the online RFC archives by BBN Corp. under the   ]
       [ direction of Alex McKenzie.                   12/96   ]
                                                                [Page 2]