Valid EAUT Email Addresses?
David Fuelling
Currently draft 5 of the spec defines an EAUT Email Identifier as
"...a "local-part" and a "domain" part, separated by an "@" sign".
By this definition, the spec is allowing email addresses in the form
"Beth Smith <be...@example.com>", which is technically a valid email
address, but not necessarily the kind of the "local-part" that we want
to accept. Instead, we're really looking for the 'dot-atom' component
of the 'local-part' in an RFC2822 email address (i.e., "beth" in the
above example).
The question is, should the final spec be liberal in defining what an
email address looks like, as draft 5 currently is....or should the
spec be specific in stating that an email address should look more
like "be...@example.com" instead.
I've codified a page (http://groups.google.com/group/eaut/web/rules-
for-valid-email-addresses) that outlines what I'm thinking, though it
deserves more debate. I'd appreciate any thoughts on this subject.
David
ps - If we leave the spec alone, we may still desire to use the rules
page I created as a "best practice" for library developers.
Will Norris
> Hey All,
>
> Currently draft 5 of the spec defines an EAUT Email Identifier as
> "...a "local-part" and a "domain" part, separated by an "@" sign".
>
> By this definition, the spec is allowing email addresses in the form
> "Beth Smith <be...@example.com>", which is technically a valid email
> address, but not necessarily the kind of the "local-part" that we want
> to accept. Instead, we're really looking for the 'dot-atom' component
> of the 'local-part' in an RFC2822 email address (i.e., "beth" in the
> above example).
>
>
> The question is, should the final spec be liberal in defining what an
> email address looks like, as draft 5 currently is....or should the
> spec be specific in stating that an email address should look more
> like "be...@example.com" instead.
I thought we had decided that the spec would explicitly require an
"addr-spec" of the form "be...@example.com". Would that wording not
invalidate "Beth Smith <be...@example.com>" ?
-will
Michael Richardson
On Jul 21, 2008, at 12:04 PM, Will Norris wrote:
>> The question is, should the final spec be liberal in defining what an
>> email address looks like, as draft 5 currently is....or should the
>> spec be specific in stating that an email address should look more
>> like "be...@example.com" instead.
>
> I thought we had decided that the spec would explicitly require an
> "addr-spec" of the form "be...@example.com". Would that wording not
> invalidate "Beth Smith <be...@example.com>" ?
Requiring the simple form, "be...@example.com," will make it *so*
*much* easier to implement. I'm also having a hard time coming up
with use cases for the longer form. So I'm definitely in favor of
keeping it short and simple.
David Fuelling
Oops. I think I mis-read RFC2822 this morning (it's not exactly user-friendly). After re-reading, you're right that "Beth Smith <be...@example.com>" would be invalid for EAUT purposes if we keep the "addr-spec" wording.
However, addr-spec allows for a "quoted-string" component, which is still kind of odd to me (e.g., RFC2822 seems to indicate that "beth smith"@example.com would be a valid addr-spec, as long as its quoted).
It's sort of a nit, I guess....but for simplicity we may want to be more granular than "addr-spec". I guess I'm not sure if there are lots of "quoted-string" or "obs-local-part" email addresses floating around (these are 2 of the 3 allowable components of an "addr-spec" in RFC2822).
Perhaps we just specify "addr-spec" in EAUT, but in implementations acknowledge that "quoted-string" and "obs-local-part" email addresses are pretty rare at this point? Not sure if that's good spec writing or not.
David