[bradleyland]

I think maybe you should take a step back and realize that the way you communicated your message initially isn't consistent with the defense you're making here. You didn't state your original issue as a bug, you called Google out as if this were some kind of feature:

"People, when you design an interface, don't put invisible, undiscoverable, irrecoverable actions in it."

This would only be a feature if it happened for everyone, and now you're using curt language with someone who is trying to help.

Troubleshooting is hard, and the communications involved are often polarizing. End users commonly feel defensive, and troubleshooters often fail to couch their language diplomatically.

The matter of reproducibility is the basic litmus test by which programmers classify issues as bugs versus environment issues. Bugs are problems in the software, and environment issues are problems in the user's specific configuration. Often times an issue is a combination of both. The willingness to engage in troubleshooting environment issues varies by developer, but if you're interested in receiving help, it's worth acknowledging that your problem falls somewhere on this continuum, which means you always bear some responsibility in remaining objective during the troubleshooting process. You seem convinced that this is some type of feature issue, rather than remaining open to other possibilities.

I think I might have an idea about what's happening in your case, but I'm reluctant to even posit the idea, because I fear you'll lash out as me as well.

[ColinWright]

I acknowledge that I was curt. I do, however, believe that I'm being consistent. It's news to me that it's perceived as not, so I'll take another look at it. Thank you.

And I didn't call it as a bug because it happened to me twice, it has happened to me on other platforms, and I assumed that a Google Mail interface would behave the way Google intended. I acknowledge that testing and debugging is hard, but I guess by default I had assumed Google would get it right. Your assertion is that they haven't got it right, and that this is a bug. Noted.

For reference, I work on software used in safety-critical situations, and so I know a little about getting user interfaces consistent, repeatable, and tested. I have the advantage that our system can control the hardware and environment, but I am aware of many of the issues. I'm just annoyed that hitting "TAB" caused me to send an incomplete message to an important customer, making me look stupid. I guess I'll have to treat Google's interfaces with the same paranoia as I use professionally when I write software. That's disappointing, but I guess unavoidable given their constraints.

And I'm sorry I've made you feel that you can't propose a suggestion as to what might have happened. That's really, really not intentional. But when I'm already having a bad morning, having someone say "No, that doesn't happen" when it clearly did is not a sure fire way of improving it.

So put it all down to the problems of writing software for uncontrollable environments causing different and sometimes undesirable behavior, coupled with the inability to read people accurately in a forum type of context, coupled in turn with a certain amount of stress over the whole thing.

And I apologize unreservedly for offence caused, and thank you for your comment.

Added in edit: I've upvoted you - I genuinely appreciate your comment.

[bradleyland]

All understandable points. I'll put out a couple of suggestions. Keep in mind that when troubleshooting, these are all guesses. They're not intended to imply that you don't have a problem. Also try to keep in mind that there may be some small detail you've overlooked that is causing the behavior you're observing. If you're dead set on insisting that the behavior happens the way you've described and there are no other possibilities, then you should stop reading here.

As troubleshooters, we often try to fill in the blanks with what we perceive as small pieces of missing information. This isn't an individual assault, but rather learned behavior. I'm damn good with a computer, but my #1 rule of troubleshooting is to assume that any "truth" about what I've observed can be invalidated through empiricism.

Enough hedging, let's see if we make any progress in troubleshooting the problem you experienced :) What we know so far:

* Pressing the TAB key results in an email being sent

* Other users report that pressing TAB doesn't send the email

We accept these as truths, but with two caveats: 1) they may be disproved or refined at any time (we aren't sentimentally attached to them, ego be damned), 2) there may be some additional information that is not yet obvious, so the list shouldn't be considered complete

What can we tease from this information?

1) Can you get the behavior to occur predictably? If you open a new compose window and press TAB, does the email send every time?

2) What is the normally expected behavior of TAB? In most web browsers, the TAB key is used to move focus from one form element to another. In Windows, the "focus" can move to buttons as well. In OS X, the focus navigation using TAB is constrained to text fields and lists by default, but can be configured to "All Controls" through the Keyboard preference pane.

3) What processes are available to intentionally send an email? That I am aware of, you can: click the send key, use the keyboard shortcut ctrl+return or cmd+enter (Windows/OS X). Use TAB to move focus to the "Send" button and press the spacebar/return/enter.

4) What small pieces of information might be missing? We've already ruled out browser extensions, so that's a good start. Try to think of any other software that uses global keyboard shortcuts or performs automation.

My pet theory (guess) would be that you're pressing TAB expecting that the TAB character is inserted in to the text field, but this is actually causing focus to move to the "Send" button. From there, a press of the spacebar, return, or enter key is sending the email. This is very easy to do when typing. If you can't regularly reproduce the behavior by a simple press of the TAB key, I'd say this gives this theory a higher probability of correctness. If you can reproduce the behavior by simply pressing TAB, then you can discard this theory.

Let me know if you have any thoughts on these items.