Whatever you do, DON'T follow this advice. It's utterly wrong.

Here's the professional way to do this:

"An application should respond with a generic error message regardless of whether the user ID or password was incorrect. It should also give no indication to the status of an existing account."

https://www.owasp.org/index.php/Authentication_Cheat_Sheet

This is a terrible idea. At this rate, you're just handing out your data! It is a matter of practicality and security. When someone enters a username and password, you do not know who may be on the other end. Telling them that the username is correct but the password is wrong puts an attacker one step closer to gaining access.

In my perfect world signing up to a service would only need my email address. I'd enter it and the service would email me a link. Clicking that link would log me in. No password necessary. My log in credentials would be available everywhere my email is available (which is everywhere), and if I accidentally delete the email then just filling in the "sign up" box with the same email address would email me a new one.

The downside is that if my email account is compromised then my account with the service is also compromised, but if that's the case then most 'forgotten password' systems would be equally broken.

This is, as everybody noticed - HORRIBLE advice. Imagine this: You mind your own business on - let's say a man-to-man datings site. You don't appear to have a problem with homosexuality, but MAYBE your wife or your employer does. Like this, when they put in your private, personal email address and the website basically tells them: "Yep, you got it, that guy has an accout here". This "meta data" is quite sensitive information, if you ask me, not only from a security point of view, but from a privacy one. Holy moly, don't follow this advice.

Everyone's already covered how this is absolutely wrong, but I wanted to add that a second query is not appropriate, either.

You'd want your query to be something like

SELECT password_hash = {password_hash} AS authenticated FROM accounts WHERE email = {email}

That way you'd be able to tell based on your result set whether the user exists, and if so if their password is valid. All with one query.

But, again, you shouldn't do this.

There's a key question whether or not to do this that I haven't seen discussed: is the fact that a given login has an account private information? Sometimes it is, sometimes it isn't.

For example, on HN, it's trivial to find out whether a given username has an account -- just visit https://news.ycombinator.com/user?id=zck , and you'll see whether zck is a registered account.

On services like that, I don't see the harm in distinguishing "password incorrect" errors from "no user" errors, since that doesn't give you private information.

But on private sites, or if you're logging in via email? That's a different story.

In a perfect world, you don't login at all because your device knows your identity and there are standardized identity federation protocols that applications and devices can participate in.

(Android already has this. Android developers: Please use AccountManager. I am highly unlikely to create an account with you.)

I can't count the number of times I've just "oh well, nevermind" when I couldn't remember the stupid credentials I used with some application, or couldn't be bothered to create new ones. Not to mention the well-documented security problems with asking users to manage passwords.

Blech I say. Blech. I'm sad that it's 2014 and this is still unsolved.

It's a good UI, however if not done right it can be potentially dangerous. As you mention in the PS, people can abuse it in various different ways. Spam bots can go around websites that use this practice and check email validity, for example.

No, in a perfect world, THIS is how web sites would handle login forms:

https://www.grc.com/sqrl/sqrl.htm

In a perfect world, username enumeration is everywhere?