Don't use regular expressions to "validate" email addresses

[John Machin]

Ben Finney wrote:
> Steve Holden <steve at holdenweb.com> writes:
>
> > Ben Finney wrote:
> > > The best advice I've seen when people ask "How do I validate
> > > whether an email address is valid?" was "Try sending mail to it".
> > >
> > That only applies if it's a likely-looking email address. If someone
> > asks me to send mail to "splurge.!#$%*&^from at thingie?><{}_)" I will
> > probably assume that it isn't worth my time trying.
>
> You, as a human, can possibly make that decision, if you don't care
> about turning away someone who *does* have such an email address. How
> can an algorithm do so? There are many valid email addresses that look
> as bizarre as the example you gave.
>
> > > The sending system's mail transport agent, not regular
> > > expressions, determines which part is the domain to send the mail
> > > to.
> > >
> > > The domain name system, not regular expressions, determines what
> > > domains are valid, and what host should receive mail for that
> > > domain.
> > >
> > > Most especially, the receiving mail system, not regular
> > > expressions, determines what local-parts are valid.
> > >
> > Nevertheless, I am *not* going to try delivery to (for example) a
> > non-local address that doesn't contain an "at@ sign.
>
> Would you try delivery to an email address that contains two or more
> "@" symbols? If not, you will be denying delivery to valid RFC2821
> addresses.
>
> This is, of course, something you're entitled to do. But you've then
> consciously chosen not to use "is the email address valid?" as your
> criterion, and the original request for such validation becomes moot.
>

What proportion of deliverable e-mail addresses have more than one @ in
them?
It may be a good idea, if the supplier of the e-mail address is a human
and is on-line, to run a plausibility check -- does it look like the
vast majority of addresses? Sure,
"fred at final.com@relay2.net at relay1.net" may be valid and deliverable,
but "clumsy at pastetwice.unorgclumsy@pastetwice.unorg" may be valid and
undeliverable. IMHO a quick "Please check and confirm" dialogue would
be warranted.

Cheers,
John