[whatwg] Comments on the definition of a valid e-mail address

[Thomas Broyer]

On Mon, Aug 31, 2009 at 7:53 AM, Ian Hickson<ian at hixie.ch> wrote:
> If we let
>
>   X = [-a-zA-Z0-9!#$%&\'*+/=?^_`{|}~]+
>
> ...then the regexp is:
>
>   ^X(\.X)*@X(\.X)*$
>
> I believe this is correct, yes.

I used to have a mail address whose local part had both a leading and
a trailing dot: .tom. at wanadoo.fr
I never had any issue with it (and I tried many mail UAs at that time,
no one ever ruled me out), except some web forms absolutely wanted (I
supposed) \w+(\.\w+)*@\w+(\.\w)*\.[a-zA-Z]{2,3} or something similar
(this is a regex commonly found on the web to validate email addresses
–at least it was something like 10 years ago–)

IIUC RFC 5322, it is explicitly allowed if written as
".tom."@wanadoo.fr (which I assume would only be true when used in
e-mail header fields; I'd personally expect a web form to accept
.tom. at wanadoo.fr, and use whatever syntax is needed –i.e. using a
quoted string for the local part in this case– when sending a mail;
I'd actually even expect this transformation to be done by the library
used to send the mail: libsmtp or whatever)

For a "good" regex to validate email addresses without too many false
negatives (false negatives are far worse than false positive re. user
experience), see http://www.regular-expressions.info/email.html

But the one that you've now put in the spec is fine by me.

-- 
Thomas Broyer