[rabarberski]

Originally Posted by BritTim

In fairness, when you select a message in the middle of the message list and subsequently delete it, it may be obvious to you which direction you are using to go through the messages, but the website cannot read your mind./

To be honest, the "smart" navigation approach I suggested isn't my own idea. I've seen it used (and enjoy using it) in various email clients.

I just checked, and indeed Apple Mail (version 9.3) has smart navigation, and it works as follows (at least from my experience, I can't know how it works under the hood):

1. Assume a list of 10 messages; let's call the top-most message number 1, and bottom-most message number 10
2. The user selects/opens a message in the middle of the list, say message number 5
3. The user presses delete
4. Since the client's (initial) default navigation direction is "down", the next message being shown is number 6
5. However, since the user actually wants to go "up" in the list, he/she presses the "up" navigation button.
6. As a result, message number 4 is now shown. Also (this is the "smart" part!), the client now sets the navigation direction as being "up".
7. The user presses delete (to delete message 4)
8. Since the navigation direction was set as "up", the next message being shown is number 3. Just like the user wants it.

So, the user has to "correct" the direction only once (by pressing the "up" button in step 5) and from then on the client continuous navigation in the same direction.
In the current Fastmail client, in step 8 the message number 6 would be shown again, requiring the user to press "up" after each delete

Of course, the smart navigation adapts when the user changes his direction (by pressing the "down" button)

A possible idea could be to go forward if the next message is unread, backwards if the previous message is unread, and navigate to the nearest unread message if neither next or previous is unread. Again, complicated/

I agree that the unread/read based algorithm would be complicated and confusing. But to our luck, we don't need it, as a simple algorithm is available :-)