Re: [sieve] Duplicate test

[Stephan Bosch <[e-mail]>]

On 1/18/2013 8:22 PM, Ned Freed wrote:
>> 2). I found that "the side effect of the test takes force at the end of
>> a successful script execution" to be odd/awkward. Maybe we can have an
>> explicit action? Or maybe leaving this as is is Ok... Need to think a
>> bit more about that.
>
> I'm not wild about this either, mostly because when you do this as two 
> steps
> there's a window where overlapping script executions can cause a 
> duplicate to
> be missed. Of course locking can be used to prevent this, but given 
> the scale
> we operate at any use of locking has to be carefully considered.
>
> That said, the problem where a script goes wonky and ends up marking 
> something
> as a duplicate is much more severe, so IMO the draft is handling this 
> the right
> way.
>
> tl;dr: Missing a duplicate is preferable to marking a message as a 
> duplicate
> incorrectly.

Exactly.

Perhaps I should mention these considerations more explicitly in the draft?

Regards,

Stephan.