My recent work on a consumer-facing product brought up the old problem: how to validate an email address before using it? There's a classic tension here between those developers that want to prevent typos from floundering around in the system, giving users feedback as soon as possible that it seems like they've made a mistake; and those developers that want to be sure that any valid email address can be used.
The usual advice on this matter is to not bother with validation, because it's a fool's errand, instead simply send an email to the address with a confirmation link. If the user clicks the link, then the address must have been valid.
I don't like this advice because the vast, vast majority of email addresses do validate with a simple regex, and the vast, vast majority of failures against the regex represent real mistakes, not obscure but valid email addresses. Catching user mistakes early is a good thing. Having the user wait for an email that will never come, then go back to enter their email address again is a pain.
This is the regex I used:
/^[^@ ]+@[^@ ]+\.[^@ ]+$/
In other words, an email address has to have stuff, at-sign, stuff, dot, stuff. The stuff can have dots in it, but can't have at-signs or spaces. And by the way, before matching against the regex, trim whitespace from the ends of the address.
As a gesture of reconciliation with the purists, I propose this: check the user-entered email address against this regex. If it matches, it's valid. If it doesn't match, show the user an "invalid email address" error box that has two buttons: "Fix mistake" which lets the user re-enter an email address, and "Use it anyway" which takes the email address as-is even though it failed the match.
This is the best of both worlds, since the common case of a catchable typo in an email address will force the user to double-check their entry, but any address can be used if the user knows what they are doing. Most users will never see the error box, since they'll enter their address correctly.
I've never seen a work flow like this, but it seems like a really simple solution to the problem. Is there something I'm over-looking? Is it too geeky?