[net.mail.headers] Invalid/questionable host names in mail

wales@ucla-locus.ARPA (Rich Wales) (09/18/85)

The following summary of invalid or questionable usage of host names in
outgoing mail is based on approximately 5,700 ARPANET messages received
at LOCUS.UCLA.EDU between 14-Aug-85 and 11-Sep-85.

I analyzed the host names used in the "return path" addresses of
incoming messages (SMTP "MAIL" commands), as well as the names which
the remote hosts used to identify themselves (SMTP "HELO" commands).
Specifically, I attempted to find two different usage patterns for
host names:

(1) Host names which did not appear in the official NIC host name table,
    HOSTS.TXT -- since addresses with such names cannot be replied to.

(2) Non-domain-style "nicknames" which appear in the NIC table.  Since
    these names are not going to be carried over to the new distributed
    domain data base, addresses with such names cannot be replied to
    from hosts which have converted their software to use the data base.

For this study, I used Version 481 of the host name table (12-Sep-85).

I would make the following recommendations based on this study.  Keep
in mind, please, that I am only a "private citizen" and do not hold any
net-wide position of responsibility; these are only suggestions, not
commands from on high.

(1) EVERY host responsible for one of the undefined return-path host
    names in Section 1 below should take IMMEDIATE steps to correct the
    situation -- either by using the correct host name in outgoing mail,
    or by having their name of choice registered with the NIC so that it
    will be added to the host name table.

    Hosts whose names are in the domain data base, but not in the NIC
    table, should consider the fact that -- whether we like it or not --
    a sizable fraction of the net has not yet converted to the domain
    data base, is not liable to do so for some time to come, and will
    thus continue for the time being to be totally dependent on the NIC
    table for name->address mappings.  You may not like this fact --
    indeed, you may consider it to be a bletcherous and utterly irre-
    sponsible state of affairs -- but it is nevertheless the way things
    are, at least right now.  Refusals "on principle" by isolated indi-
    vidual hosts to deal with those that are still using the NIC table
    is unlikely to accomplish anything (except to make it harder for
    people to send mail to each other).

(2) EVERY host responsible for one of the non-domain-style nicknames in
    Section 2 below should convert over to the appropriate domain-style
    name as soon as possible.

    Since NO non-domain-style nicknames (i.e., host names without at
    least one period) will EVER be listed in the domain data base, these
    nicknames -- however near and dear they may be to the hearts of
    their users -- will slowly become useless as more and more hosts
    wean themselves from the NIC table and convert over to the domain
    data base.  No ifs, ands, or buts, folks -- this is REALITY.

(3) Hosts listed in Sections 3 and 4 below should take steps to make
    their "SMTP sender" or "net mailer" programs use a correct, domain-
    style host name in the SMTP HELO command.  As in the case of return
    addresses, it is probably advisable (for the time being, at least)
    that new domain-style names be registered in the NIC table as well
    as in the domain data base.

    Although most SMTP servers in use today make no serious attempt to
    validate the HELO argument, some people have proposed to check the
    sending host's name and to reject the HELO command (thus blocking
    mail transfer) if the name is unknown.  It therefore makes sense to
    ensure that the correct host name appears in the HELO.

(4) Hosts who have converted over to using the domain data base may want
    to consider modifying their name->address mapping routines to try
    adding ".ARPA" to the end of a non-domain-style host name if it can-
    not be found in their tables "as is".  (I.e.:  if presented with the
    host name "PODUNK", try looking up "PODUNK.ARPA" before giving up.)
    This will let you make sense of most (though unfortunately not all)
    non-domain-style nicknames.

    I suggest this (admittedly disgusting) "hack" only because Murphy's
    Law dictates that lots of hosts will probably continue to use their
    old non-domain-style nicknames until the bitter end -- and maybe
    even beyond.  Again, our first priority should hopefully be to max-
    imize the probabilities that mail will get through -- since large
    numbers of users on many systems are mail-naive in the extreme and
    will have absolutely not a clue as to what alternatives to try if
    the return address on a piece of incoming mail can't be replied to.

Since much of the following summary was processed by hand, it may con-
tain some minor errors.  I take full responsibility for these and apolo-
gize in advance to any hosts which I might have inadvertently pointed an
accusing finger at.

Also, I realize that many hosts listed below may already be aware of the
fact that they are using an incorrect host name in their mail -- and are
either at work on the problem, or are looking for a solution.  Or, some
hosts may have been using an incorrect name until very recently, but
have now fixed things.  My intent in posting this list is not to badger
or criticize anyone; I am simply trying to help make sure everyone is
aware of the problem and is doing whatever they can to resolve it.

Further, let me emphasize that this summary covers only those hosts
which have sent mail to LOCUS.UCLA.EDU during the past month.  Just
because a host is not listed does not necessarily mean that they are
doing the right thing with regard to host names in mail.

Since many of the people who most need to read this message probably do
not subscribe to this list, I will also be sending individual messages
to "postmaster" at each affected host (where the identity of said host
can be determined).  My apologies in advance to anyone who thereby gets
a "double whammy" (or even a "triple whammy") from me on this subject.

Unfortunately, I am not able to offer any specific help to anyone who is
trying to fix their mail software.  In particular, I am not a "sendmail"
guru, and I do not have any fixes to "sendmail" which might be required
in order to make it conform to current name requirements.  If anyone has
devised such fixes to various widely used mail systems, it might be nice
if said fixes (or pointers to them) could be posted.

========================================================================

(1) The host name in the return-path address (SMTP MAIL command) is not
    in the NIC table.

	Return-Path Host Name    Host Sending Mail
	    EAERO2                   AERO2.ARPA
	    EAR.BERKELEY.EDU         UCB-VAX.BERKELEY.EDU
	    HOTDOG.ARPA              DECWRL.ARPA
	    IC.BERKELEY.EDU          UCB-VAX.BERKELEY.EDU
	    IPTO-VAX.ARPA            IPTO.ARPA
	    ISI-DAWN.ARPA.ARPA       ISI-DAWN.ARPA
	    MIRO.BERKELEY.EDU        SEISMO.CSS.GOV
	    MIT-BORAX.MIT.EDU        MORDRED.PURDUE.EDU,
				     SEISMO.CSS.GOV
	    MIT-CLIO.ARPA            SEISMO.CSS.GOV
	    PLAYFAIR                 36.10.0.171
	    SALLY.LOCAL              SALLY.UTEXAS.EDU
	    TP4                      192.5.14.154

    NOTES:
    
    (a) The names {EAR,IC,MIRO}.BERKELEY.EDU and MIT-BORAX.MIT.EDU do
	appear in the domain data base (but not in the NIC table).
	Hence, only hosts which have converted over to the domain data
	base can send mail to these places.  The wizards at these
	places should seriously consider having these names listed in
	the NIC table, as well as in the domain data base.

    (b) In certain cases above, the "host sending mail" may have been
	acting as a relay for the actual originating host.  Hence, the
	"host sending mail" is not necessarily the correct, official
	substitute for the undefined "return-path host name".

    (c) I have nary a clue as to the identity of "HOTDOG.ARPA".  The
	full address was "sci!kuo@HOTDOG.ARPA".  Ideas, anyone?

(2) The host name in the return-path address (SMTP MAIL command) is
    listed in the NIC table -- but it is a non-domain-style nickname
    (and thus will never be in the domain data base).

	AERO2          GE-CRD         MITRE-BEDFORD  SRI-UNIX
	AERO3          GLACIER        NLM-VAX        SU-STAR
	AIDS-UNIX      GREGORIO       NRL-CSS        TALCOTT
	BBN-UNIX       IPTO-VAX       NTA-VAX        TOVE
	BBNCCV         ISI-ELVIRA     NYU-CMCL2      UCI-ICSA
	BERKELEY       ISI-HOBGOBLIN  OMNILAX        UCI-ICSD
	CCA-UNIX       ISI-VAXA       ORNL-MSR       UCI-ICSE
	CIT-750        JPL-MILVAX     PESCADERO      WHITNEY
	CIT-VAX        LLL-CRG        ROCKEFELLER    YALE-APVAX
	DIABLO         MARYLAND       SAFE
	EDN-VAX        MIT-EDDIE      SRI-SPAM

    NOTES:

    (a) Berkeley was using domain-style names for a time, but was forced
	to switch back to using "Berkeley" for the time being due to
	some kind of software problem.  The last word I had from them
	was that they were hoping to be using domain-style names again
	in another month or so.

    (b) Most of the above names can be converted to valid domain-style
	host names simply by adding ".ARPA".  Exceptions to this rule
	are the Stanford hosts DIABLO, GLACIER, GREGORIO, PESCADERO,
	SAFE, and WHITNEY (whose official domain-style names are,
	respectively, SU-AIMVAX.ARPA, SU-GLACIER.ARPA, SU-GREGORIO.ARPA,
	SU-PESCADERO.ARPA, SU-SAFE.ARPA, and SU-WHITNEY.ARPA); IPTO-VAX,
	whose official domain-style name is IPTO.ARPA; and NYU-CMCL2,
	whose official name is NYU.ARPA.

    (c) Lots of messages relayed through mailing lists on YALE.ARPA
	showed non-domain-style nicknames in return-path addresses --
	even though mail arriving directly from the hosts in question
	used the official names.  I suspect the YALE.ARPA software may
	be rewriting the addresses as the mail is relayed through.  I
	did not include any mail relayed through YALE.ARPA when putting
	together the above list on non-domain-style nicknames.

(3) The host name given in the SMTP HELO command is not defined in the
    NIC table.  (Some of the hosts in this category are not in the NIC
    table at all; one is listed only as a gateway.)

    The "official name/address" info below is derived from the Internet
    address of the host which actually made the SMTP connection.

	Official Name/Address    Name in HELO Command
	    AERO2.ARPA               EAERO2.ARPA
	    AIDS-UNIX.ARPA           AIDS-GRAPE.ARPA
	    IPTO.ARPA                IPTO-VAX.ARPA
	    ISI-DAWN.ARPA            ISI-DAWN.ARPA.ARPA
	    MIMSY.UMD.EDU            MARYLAND.ARPA.ARPA
	    MIT-MC.ARPA              MIT-OZ
	    MORDRED.PURDUE.EDU       MORDRED.ARPA
	    NYU.ARPA                 NYU-CMCL2.ARPA
	    OMNILAX.ARPA             OMNILAX.UUCP
	    RIACS.ARPA               RIACS-GW.ARPA
	    THINK.COM                GODOT.THINK.COM
	    UCSF-CGL.ARPA            UCSF-CGL.UCSF
	    USC-OBERON.ARPA          OBERON.ARPA
	    YALE-COMIX.ARPA          YALE-COMIX.YALE.ARPA
	    YALE-MTRANS.ARPA         YALE-MTRANS.YALE.ARPA
	    36.10.0.171              PLAYFAIR.ARPA
	    192.5.14.154             TP4.UUCP

    NOTES:

    (a) Despite their domain-like appearance, YALE-COMIX.YALE.ARPA and
	YALE-MTRANS.YALE.ARPA do not appear in the domain data base.

    (b) The name GODOT.THINK.COM is in the domain data base, but not in
	the NIC table.

    (d) As far as I can tell, MIT-MC.ARPA calls itself "MIT-MC.ARPA"
	most of the time.  We recorded only two instances in which it
	called itself "MIT-OZ"; both times, it was acting as a mail
	relay for MIT-OZ (an internal MIT site, not on the Internet).

(4) The host name in the SMTP HELO command is listed in the NIC table --
    but it is a non-domain-style nickname (and thus will never be in the
    domain data base).

    As above, the "official name" info below is derived from the Inter-
    net address of the host which actually made the SMTP connection.

	Official Name            Name in HELO Command
	    AEROSPACE.ARPA           AEROSPACE
	    AMSAA.ARPA               AMSAA
	    BBNCCM.ARPA              BBNCCM
	    CSNET-RELAY.ARPA         CSNET-RELAY
	    IPTO.ARPA                IPTO-VAX
	    SU-AIMVAX.ARPA           DIABLO
	    SU-GLACIER.ARPA          GLACIER
	    SU-GREGORIO.ARPA         GREGORIO
	    MITRE-BEDFORD.ARPA       MITRE-BEDFORD
	    NRL-CSS.ARPA             NRL-CSS
	    SU-PESCADERO.ARPA        PESCADERO
	    SU-SAFE.ARPA             SAFE
	    TYCHO.ARPA               TYCHO
	    UCI-ICSA.ARPA            UCI-ICSA
	    UCI-ICSD.ARPA            UCI-ICSD
	    UCI-ICSE.ARPA            UCI-ICSE
	    UCL-CS.ARPA              UCL-CS
	    SU-WHITNEY.ARPA          WHITNEY
	    YALE.ARPA                YALE

========================================================================

Rich Wales // UCLA Computer Science Department // +1 213-825-5683
	3531 Boelter Hall // Los Angeles, California 90024 // USA
	ARPA:   wales@LOCUS.UCLA.EDU  -or-  wales@UCLA-LOCUS.ARPA
	UUCP:   ...!(ucbvax,ihnp4)!ucla-cs!wales