Inconsistent interface with non-bounced email

[saghaulor]

@mitio I noted some inconsistencies in the interface of the gem. I wasn't sure how you might want to handle it, if at all. I was using the following file: https://github.com/livebg/bounce_email/blob/master/test/non_bounces/tt_1234210666.txt

I would expect the behavior to be:

bounced_mail.bounced? => false
bounced_mail.type =>  'Success'
bounced_mail.code => '99'
bounced_mail.reason => 'Vacation auto-reply'

However, the behavior is:

bounced_mail.bounced? =>  false 
bounced_mail.type => 'Permanent Failure'
bounced_mail.code => 'unknown' 
bounced_mail.reason => 'unknown'

I've started work to make it more consistent, but if you don't think we should change anything, I'll abandon the work. Please advise.

[saghaulor]

Part of the change that I started to implement was to change the code to '2.9.9', because the '2' prefix indicates a successful transmission, as per https://tools.ietf.org/html/rfc3463.

[mitio]

Well, this definitely is not a permanent failure.

Your proposed response to this particular auto-reply email looks sane.

However, I have no idea how to detect auto-replies. Do you have any suggestions on that?

What I'd be reluctant to do is consider all unknown emails as auto-replies and tag them as "success".

[saghaulor]

@mitio Well in the particular case of the above example, we can determine that it's an out of office email based on the body content. However, for so reason it isn't recognized as such. The work I started would still have unknown, but it would only apply to those that are truly unknown.

The question then is, is a boolean appropriate as an indicator to determine if it's a bounce or not? It seems to me that there are three states, bounced, not bounced, and unknown. I'm not sure if we should change this part of the interface, but not doing so forces us to decide, is an unknown mail bounced or not bounced? To maintain the current behavior, an unknown email would be not bounced.

It sounds like you're open to improving the interface. I'll keep working on it then and push my changes. Then you can decide if it's appropriate or not, as well as propose your astute suggestions.

[mitio]

I'm thinking we could have another boolean, e.g. recognized? and when recognized? is false, then bounced? could return nil. This will be mostly backwards-compatible.

I am open to improving anything :) So please go ahead.