Web-development – Who should write the test plan

In my acceptance suites I have stayed away from using technology specific controls i.e for web applications don't use css dont use html elements if you need to fill in a form do the specifics in the steps to setup the SUT not the actual acceptance tests

I use cucumber for my acceptance and have the following

Given A xxx 
And I am on the xxx page
And a clear email queue
And I should see "Total Payable xxxx"
And I supply my credit card details
When I the payment has been processed
Then my purchase should be complete
And I should receive an email
When I open the email with subject "xxx"
Then I should see the email delivered from "xx"
And there should be an attachment of type "application/pdf"
And attachment 1 should be named "xxxx"
And I should be on the xxx page
And I should see my receipt

this example is back by a web application but I can still use the test to test against a desktop application as the steps are used to setup the SUT not the acceptance tests

this test sits at the end of a purchase which goes

Generate -> Confirm -> Payment -> Print Receipt

the test above is for the payment step the other steps are setup in other tests due to the application being able to setup into these states with data or http actions in this case the payment has a given which does the confirm steps and the confirm does the generate steps so they are a bit brittle at the minute

yes, there can be duplication. Often with line of business type websites, you find the acceptance tests can actually substitute for developer oriented tests, because the depth of code needed to implement features is not too great. As a comparison, image recognition software may have huge depth, but at an acceptance level it might be simple as needs to recognize this as a circle, and this as a square. Developer wise, you need to test your image capture, light correction, object detection, object recognition, blah blah....

So if you have a shallow amount of code between acceptance and implementation, it quite often does feel redundant.

However, there are a few other differences in terms of process. You can have a suite of acceptance tests build up, and they don't all have to pass all the time, they can act as targets, so you go from 5% passed, to 10% to 50% etc, until you knock all the tests off. In some cases, ideally these tests can be put together semi independently of the developer. Developer oriented tests, the intention there is to have them passing all the time.