NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/sys/sgi/12678

Internet Message Format | 1992-08-20 | 7.1 KB

Path: sparky!uunet!cs.utexas.edu!sdd.hp.com!uakari.primate.wisc.edu!ames!sgi!rhyolite!vjs
From: [e-mail] (Vernon Schryver)
Newsgroups: comp.sys.sgi
Subject: Re: 4.0.5 and mail
Message-ID: <[e-mail]>
Date: 20 Aug 92 17:35:47 GMT
References: <[e-mail]>
Organization: Silicon Graphics, Inc.  Mountain View, CA
Lines: 138

In article <[e-mail]>, [e-mail] (Timo Suhonen) writes:
>
> Just got 4.0.5 and installed it. Release Notes says that IRIX is now
> 8-bit system with Latin-1 as default. Why does mail clear the 8:th bit?
>
> I know that RFCs allow mail to do it, but here in Finland I'd like
> to use the Latin-1 character set in my mails. The international mails
> also seems to accept 8-bit. At least mails to HP at California (?) and
> mails to Iceland survives full 8-bit.
>
> I have tried to compile the IDA-sendmail, but I had only two or tree
> hours time to do it and the compilation was not just "plug in and make"
> -system. I have to return to IDA-sendmail some time later if SGI keeps
> clearing the 8th bit.


Some people are spreading false statements about the effects and
effectiveness of versions of sendmail that do not clear the 8th bit of
text or headers.  There has recently been some vigorous "discussion" in
the main IETF mailing list between those working on the extensions to
SMTP (the mail transport standard) and the "just send 8bits" crowd.

The facts of the matter are these, some important and some not:

    1. RFC-822 requires that the 8th bit be cleared on all mail and headers.
    It is not a matter of "allowed," but of "required."

    2. standard BSD sendmail has cleared the 8th bit for years.  This
    means there are 10's of thousands of 8-bit clearing mail hosts
    and relays directly connected to the Internet.

    3. it is easy to build sendmail so that it does not clear the 8th
    bit.  Such a sendmail with Mail User Agents that generate
    8-bit-unclean text and headers makes the system non-comformant
    with current standards.

    4. such a non-conformant sendmail cannot reliably deliver 8-bit email.

      (a) the writer above is wrong about mail between Finland and
    California surviving with its 8th bit.  It is no doubt true
    that mail survives almost all of the time.  However, it is
    practically impossible for most people to ensure that no
    7-bit-stripping mail relay will ever see and so strip your
    text.  All it takes is a temporarily unreachable destination
    for your 8-bit email to use an MX record that causes your mail
    to be relayed through a 7-bit site.  The result is that your
    mail will be occassionally corrupted.  It will sometimes arrive
    at its destination with the 8th bits missing.

      (b) If you use 8-bit characters in a From: or To: line and your
    message happens to pass through an old mail relay (as in #a
    above), then your addresses will suddenly become nonsense.
    Your mail message will not be deliverable nor returnable.
    Having your mail sometimes disappear is worse than having
    it delivered as nonsense.

      (c) plenty of Internet mail goes through UUCP links.  Does it make
    any sense to expect that mail to have its 8th bit preserved?
    #a and #b apply in spades for the majority of hosts reachable
    by Internet mail, since they have UUCP links.  (I think
    the growth of the Internet is still less than the growth
    of the encompassing UUCP network.)

      (d) there is no way for the sending and receiving machines
    to know they are using the same 8-bit character set.  I
    understand there is more than one such character set, and that
    the character set used by the "just send 8bit" crowd is not
    appropriate for Asia, no matter how fine it is for Europe and
    the Americas.  In other words, "just send 8bits" works
    for small parts of the world, but is not an international
    solution.

      (e) Sgi.com relays a substantial amount of mail.  The sendmail
    binary on sgi.com bounces any message with bytes in headers
    that look like control characters, regardless of the 8th bit.
    Those bytes crash sendmail itself, because sendmail uses
    "illegal" characters as flags and tokens.  All known sendmail
    binaries share this vulnerability.  The Northern European 8-bit
    character sets do not use such bytes, but other character sets
    do. We continue to see a steady stream of bounces at sgi.com
    caused by headers with such poison bytes.  The idea that
    mail transports are "8-bit clean" has caused many people
    to try to use character sets with graphics that look like
    control characters.

    5. It seems likely that just sending 8bits will not be the
    extension of SMTP that will be chosen by the Internet community
    in persons of the IAB and ISEG.  This is what I'm told by
    members of the so called Internet Good Ol' Boys network.  There
    have been calls for (I think unwise) explicit bannings of "just
    send 8bit" systems.

    6.  There are rumors based on statements in the IETF working group
    that vendors currently shipping "just send 8bit" sendmail will
    stop.  (I'm a little skeptical that will happen.)

    7. It will be at least 5 and probably 10 years before many of the
    machines in the Internet are upgraded to whatever SMTP
    replacement, whatever it is.  One of the most irrititating
    parts of the "just send 8bits" claims that most of the hosts in
    the Internet will be upgraded in the next few months.  It has
    been 5 or 6 years since funding agencies, authorizing
    authorities, and everyone else has absolutely, positively
    required all Internet sites to use DNS.  However, we at sgi.com
    continue to see a steady stream of requests from people at
    sites not using DNS for the IP number of sgi.com.  Claims that
    people will replace their apparently working mail systems
    sooner than they have replaced their not-working hostname
    systems is incredible.  Some of those making the claim have
    been around long enough to know better.

    8. There are Internet standards which allow sending not just
    8-bit text but arbitrary binaries, including voice and
    video.  The MIME standards are easily implemented, are
    "approved", and solve the problem.


Personally, I don't get excited about conforming or not to standards.
All that  matters to me is whether things work.  The simple fact is
that "just send 8bits" does not and cannot work.  It does not ensure
that what the author wrote arrives at the reader's screen.  On the
other hand, there are existing standards and implementations to do
work.  In the absense of the dishonesty and parochialism, the choice
would be clear and not controversial.

I think that Silicon Graphics should make available a broken sendmail
that passes 8-bits, as well as a MIME system.  If I'm right, eventaully
users will figure out that "just send 8bits" is a fraud.  However,
neither of those can be made available immediately, even if I convince
everyone necessary.  They must wait for a release to ride.

The people in the Internet community who are telling people to "just
send 8bits" are knowingly ignoring the facts above.  They are knowingly
causing people to think things they know are false.  They are
dishonest, and should be dealt with as such.


Vernon Schryver,  [e-mail]