[comp.mail.headers] Are underbars considered bad form for addresses in RFC822 headers?

ron@topaz.rutgers.edu (Ron Natalie) (04/02/88)

RFC822 which is the specification used for headers in Internet and
USENET messages (but not necessarily UUCP mail) allows _ in the local part
(to the left of the @).  Underscore is a legal charater in an atom and
is not required to be quoted (as certain other characters such as period).

-Ron

cfe+@andrew.cmu.edu (Craig F. Everhart) (04/04/88)

RFC822 treats ``_'', and several other several punctuation characters, in
exactly the same way as it treats, say, alphabetics.  These include:
        !#$%&'*+-/=?^_`{|}~

As to whether a particular mail system will recognize mail addressed to
Firstname_Lastname (or, for that matter, ``Firstname.Lastname''), that's up to
the particular mail system.  Andrew.cmu.edu, for instance, will get mail to me
no matter whether it's addressed as any of the following:
        Craig.Everhart@andrew.cmu.edu
        Craig_Everhart@andrew.cmu.edu
        Craig_F._Everhart@andrew.cmu.edu
        cfe+@andrew.cmu.edu
plus a bunch of less-reliable alternatives that presume the absence of
conflicting names, e.g.:
        cfe@andrew.cmu.edu
        Creg.Everhart@andrew.cmu.edu
        c.f.everhart@andrew.cmu.edu

per@erix.UUCP (Per Hedeland) (04/11/88)

In article <Apr.1.17.35.54.1988.3117@topaz.rutgers.edu> ron@topaz.rutgers.edu (Ron Natalie) writes:
>  Underscore is a legal charater in an atom and
>is not required to be quoted (as certain other characters such as period).

Where does this notion that periods are not allowed in unquoted local-parts
come from (I've encountered it in other places as well)? As far as I can
understand, RFC822 quite to the contrary explicitly allows this, giving

     local-part  =  word *("." word)             ; uninterpreted
                                                 ; case-preserved

as the definition. This is even discussed in some detail in para 6.2.4-5.

Regards
--Per Hedeland
per@erix.ericsson.se  or  ...{mcvax,uunet}!enea!erix!per

cfe+@andrew.cmu.edu (Craig F. Everhart) (04/14/88)

Excerpts from: 11-Apr-88 Re: Are underbars considere.. Per Hedeland@erix.UUCP
(735)
> Where does this notion that periods are not allowed in unquoted local-parts
> come from (I've encountered it in other places as well)? As far as I can
> understand, RFC822 quite to the contrary explicitly allows this...

First of all, Ron Natalie's message said that periods were illegal in
``atom''s, not in ``local-part''s.  He is correct, as are you, but he was
talking about an RFC822 atom, and you're talking about an RFC822 local-part.

One can also confuse ``phrase'' with ``local-part''.  For obscure (perhaps
mistaken) reasons, RFC822 allows (unquoted) periods in local-parts but not in
phrases, and allows (unquoted) spaces in phrases but not in local parts.
``phrase'' is what's to the left of an angle-bracket address in a header;
local-part is what's immediately to the left of an at-sign.  Thus,
        To: Craig Everhart <Craig.Everhart@andrew.cmu.edu>
is a legal header on both counts, but
        To: Craig F. Everhart <Craig Everhart@andrew.cmu.edu>
is illegal on both.

There were a lot of inertia-based reasons for disallowing spaces in a
local-part.  Disallowing periods in a phrase is probably an oversight, but
we're stuck with it.  Another curiosity is that a phrase can't be null in a
message header, so the header
        To: <Craig.Everhart@andrew.cmu.edu>
is, strictly speaking, illegal.  (When we have to use a return-path (which, if
it's a source-route, can require the ``<'' and ``>'' characters) to compose an
error message, sometimes we have to generate a non-null phrase to precede them.
 I used to use the text constant ``dummy'', representing a syntactic dummy, but
when people started to take it as a mild insult, I had to change it to
``Message sender''!)

                Craig Everhart
                Andrew message system