On pe, 2010-06-11 at 15:56 +0200, Thomas Hummel wrote:
The thread you pointed insisted on the fact that the list "" * command should not cross namespaces boudaries.
Right.
So how can a client list mailboxes of a namespace if it doesn't discover the namespace name first ?
It couldn't.
That's why I thought :
. a modern client which supports NAMESPACE, will discover the namespace name (hidden=no) and list it's content (list=yes).
Or possibly list its contents twice, in two namespaces..
But maybe you mean that dovecot list the namespace content anyway if list = yes without having the need for the client to know anything about the namespace ? In that case, wouldn't it break the "list don't cross namespace boundaries" ?
Yes. hidden=no, list=yes configuration breaks the "list doesn't cross namespace boundaries". That's why I said it's arguably wrong to use such configuration.
With hidden=no, list=no it works the way namespaces were intended to work, but few clients show the namespace's contents then. That's why Mark was talking about his chimera namespace (which is similar to Dovecot's hidden=yes, list=yes namespaces) and the idea of maybe getting rid of namespaces completely.