When a resource is temporarily unavailable, it would be handy if sendmail indicated that unavailability when the database lookup fails. Consider NFS, for example. It can time out when a server is down briefly, but a failed lookup of a user's login name need not cause a permanent failure under such a circumstance. Instead, something should be returned to show that it is only a temporary failure.

The -T database switch was added with V8.10 sendmail to solve this problem. You use it to define a suffix to add to the key for the returned failure value when the problem is temporary. You might use it like this:

Kmailservers nis -T.Defer -o mailservers
...
R $* <@ $+ > $*                 $: $1<@$2>$3 <$(mailservers $2 $: Fail $)>
R $* <@ $+ > $* <$* . Defer>    $# error $@ 4.2.2 $: "450 defer"    handle failure here
R $* <@ $+ > $* <Fail>          $# error $@ 5.7.1 $: "550 reject"   handle failure here
R $* <@ $+ > $* <$+>            $# smtp $@ $4 $: $1 < @ $2 > $3 OK, so send it
...

Note that a permanent failure returns the failure alternative indicated by the $: operator (the Fail). But a temporary failure returns the suffix defined by the -T, appended to the original key (the $2), to form $2.Defer.

Note that this definition of temporary failure is different from that defined by the -D database switch. With -D, database lookups are not done at all if the DeliveryMode option (DeliveryMode) is set to defer. Also note that this -T database switch affects only the return value. It does not affect the outcome of mail delivery. To affect the outcome on temporary failures, use the -t switch (-t).