Comments on on-line host name service

Network Working Group                                   Mark Krilanovich
RFC # 623                                                           UCSB
NIC # 22004                                            February 22, 1974
Reference: RFC #606, 608


                 Comments on On-Line Host Name Service


    Peter Deutsch in RFC #606 pointed out the desirability of having a
single host maintain a data base containing official host names and host
addresses, as well as other information of secondary importance.  Mike
Kudlick in RFC #608 agreed with the concept, and proposed that the NIC
would implement Peter's ideas.  I would like to add my voice to those in
support of such a service, and express a few ideas for its modification.

    The notion of having a single host maintain this data base clearly
has the weakness that anyone wishing to obtain a copy of the data may be
faced with the situation that the serving host is not available when the
data is desired.  It is true that each host could save a copy of the
most recently obtained data, such that whenever a current copy cannot be
obtained, at least a very recent copy is available.  This is not a
particularly attractive idea, since it requires a non-trivial amount of
bother on the part of everyone.  Therefore, I propose that the NIC
maintain the master data base, and one other host be responsible for
maintaining a secondary copy, which is to be updated to be equal to the
NIC's at periodic and often intervals, such as once a day.  This way,
anyone wishing to obtain the data can first try the NIC, and if that
fails, try the secondary host, thus much reducing the probability that
the data cannot be obtained, while requiring additional software to be
written at only one additional host.  Further, I volunteer UCSB to be
that secondary host.

    The proposal currently underway calls for the host names data base
to have the format of ASCII file.  RFC 606 makes the point, with which I
completely agree, that this data base should be formatted in an easily
machine-readable form.  To this end, I propose that the data base be
retrievable in binary form rather than ASCII.  Using this concept, for
example, <host-address> would be a one-byte (eight-bit) binary number,
<host-name> would be a one-byte length field followed by that many ASCII
characters, and the possible <attribute-values>'s for the STATUS
<attribute-name> would be one-byte binary numbers.  This modification
would clearly make the data base unintelligible to a human user, and,
just as clearly, much more easily interpreted by a program.

    RFC 608 states that the data base will be maintained as a file and
retrievable through FTP.  I question the wisdom of basing such a simple
process as keeping a host table up to date on such a complex protocol as

FTP.  Therefore I propose that the data base be available via a program
running under its own socket at the NIC and at the secondary host.  This
also avoids the necessity for the accessing program to know the login
parameters for the guest account at the serving host, which in fact
might not be the same at the two hosts.  Again, the motivation is to
make things easy for accessing programs.

    Anyone with comments about any of the above is encouraged to make
them known.










       [ This RFC was put into machine readable form for entry ]
       [ into the online RFC archives by Alex McKenzie with    ]
       [ support from GTE, formerly BBN Corp.            10/99 ]