An important question related to error recovery is that of data naming. Applications and ALMI require a commonly understood naming convention so that they can communicate which data is requested. Since losses in ALMI group are more likely to occur in batches over dispersed time intervals rather than isolated packets on regular time intervals, sequence numbers as used by TCP, are insufficient to specify a member's data reception state and could hinder a members' ability to request and retransmit data efficiently. Furthermore, an application may decide to ignore certain packets, for example, packets containing out-of-date information, and only recover others. A data naming component is thus more desirable since it allows flexibility in tailoring application reliability semantics.
In ALMI's data naming interface, an application can specify the
mapping between its application data units and ALMI packet
sequence numbers. An ADU is solely defined by application
protocol, for example, for some database applications, it can be an
object ID; or for a ftp application, a tuple containing 
file name,
offset, length
. Other more sophisticated mechanisms such as
hierarchical data naming schemes [15,5] can be incorporated
as well, to achieve better flexibility and efficiency.