Tony Hansen wrote:
I'd like to reconfirm the consensus between these two choices:
(i) change "string" to "atom"
(ii) change "string" to "dot-string"
With the examples of <Dot-string>s shown here I tend to (II).
In <http://permalink.gmane.org/gmane.ietf.smtp/5496> I proposed
to replace 2821 <Dot-string> by 2822 <dot-atom-text> because
they are syntactically equivalent.
There are only six uses of <Atom> in 2821bis-01, two of them in
the <Dot-string>. Confusingly a 2821 <Atom> isn't the same as
a 2821 <atom>, the latter is [CFWS] 1*atext [CFWS], but in 2821
it's only 1*atext.
I propose to replace the remaining four 2821 <Atom> by 1*atext
eliminating this potential cause of confusion:
- String = Atom / Quoted-string
+ String = 1*atext / Quoted-string
- Addtl-Link = Atom
+ Addtl-Link = 1*atext
- Attdl-Protocol = Atom
+ Attdl-Protocol = 1*atext
- Argument = Atom
+ Argument = 1*atext
The uses of <String> in 2821bis, ignoring the settled "ID" case,
have to be checked: Are the arguments for EXPN, VRFY, HELP,
and NOOP really a <Quoted-string> if they're no <Atom> ?
In other words, is 'HELP me now' a syntax error, should it be
'HELP "me now"' ? Comparing it with RFC 821, apparently these
strings used to be *_unquoted_* strings of any <c> (= <atext>)
or "\" <x> (= <quoted-pair>). Is 'HELP me\ now' acceptable ?
RFC 821 allowed no argument for NOOP. Please add this to the
todo list for an "implementation and interoperability report":
Are NOOP arguments supported, or do they cause havoc ?
Frank