At 4:44 PM +0100 2006-08-08, Ian Eiloart wrote:
But, the idea is NOT to try to parse bounce *messages*, it's to parse bounce *codes*.
Here's the deal.
You think it's going to be trivially easy to add this new feature, and to parse the codes correctly, with the correct outcome, all you have to do is follow the RFC and everything will be hunky-dory.
I think that the problem is a lot more complex than that, with many sites giving totally inappapropriate response codes for the real underlying reason, and trying to parse them is likely to cause more problems than it solves. Moreover, I think this is going to add unneeded complexity to the system for what I believe will be, at best, relatively minimal benefit. Worse, in order to adhere to the spirit of this idea and make the concept actually work, we'll have to get into trying to parse the actual wording of the error messages, and then we'll have to get into internationalization issues of all those words we're trying to parse, because I'm pretty sure that words like "virus" are not the same in Polish, Chinese, Farsi, and whatever other various languages we have to support.
So, here's the solution. You go implement the code to do what you're talking about, and see how it works on your site. Make sure to collect all the bounce messages in question, and the action that was taken by the system. This way, we humans can compare the performance of your new code. Once you're done tweaking the system to work as well as you can manage, come back to us and show us your code and your input data, and prove to us how well it works.
But without a patch and a strong indication that this is a significant improvement for relatively little added complexity, I don't think you're going to get any further traction in this issue.
That's it. I've said my piece. Unless you have something new to add to the discussion, I'd suggest you do us all a favour and let this drop.