"... In PAGE 5: ... RELIABLE PERIODIC BROADCAST In this section we develop optimized periodic broadcast protocols that (1) assume a maximum aggregate transmis- sion rate to any given client that is a tunable parameter, n, which may be less than (or greater than) twice the me- dia play rate, and (2) enable clients with heterogeneous loss probability to recover from packet loss. The notation used in developing the RPB protocols is given in Table2 . To deal with the fundamental challenges involved in designing such optimized protocols, we proceed in four stages.... ..."

"... In PAGE 5: ... n = 32, Binitial = min(B; f + 1). Table1 shows the measured reliability of 10,000 broadcasts among n = 32 processors for di erent values of B, F and f, where f is the number of crashed processors. No link failures occurred during these broadcasts and all processors that crashed did so before the rst broadcast.... In PAGE 5: ... This is because a processor will have a more accurate idea of which processors already have m the later it forward m. Table 2 shows the average number of messages sent per broadcast for the runs that were used to generate Table1 . The number of messages sent is bounded by BF (n ? f).... ..."

"... In PAGE 19: ... This reinforces the view that Internet users are not just more trusting in institutions generally, although broadband users were found by OxIS to be somewhat more trusting than non-users. For example, broadband users are slightly more confident in the reliability of information published in newspapers than are non-users and slightly more confident in the reliability of information broadcast on television as well ( Table7 ), but these differences are quite minor and not statistically significant (Appendix 3). Broadband and narrowband Internet users are, however, distinctive primarily in the reliability they assign to information on the Internet, which they rate almost as highly as they rate information on television, and higher than they rate newspapers (Table 7).... In PAGE 51: ... In turn, Narrowband users were more likely to give a substantive response than Past users, who were more likely to do so than Non-users. By contrast, on questions that do not relate to the Internet, Internet users and Non- users show similar levels of trust, as Table7 showed. ... ..."

"... In PAGE 5: ... Belowwegive reliability results obtained from simulations when failures are considered. Table1 shows the measured reliability of 10,000 broadcasts among n = 32 processors for di erent values of B, F and f, where f is the number of crashed processors. No link failures occurred during these broadcasts and all processors that crashed did so before the rst broadcast.... In PAGE 5: ... This is because a processor will have a more accurate idea of which processors already have m the later it forward m. Table 2 shows the average number of messages sent per broadcast for the runs that were used to generate Table1 . The number of messages sent is bounded by BF(n ; f).... ..."

"... In PAGE 2: ... Reliable transport protocols can be further subdivided into upstream (mostly unicast/convergecast transmission to aggregator or sink nodes) and downstream (mostly multicast/broadcast of code or configuration updates from sink nodes to sensor nodes). Table1 presents a taxonomy of reliable transport protocols for WSNs. It is a well established fact that providing some reliability support at intermediate nodes (hop-by-hop in the extreme) is more energy-efficient than treating reliability end-to-end only.... ..."

"... In PAGE 19: ...one indicates how reliable the broadcast is, while the second one measures the number of unnecessary transmissions. Table1 shows the above two measures averaged out for the whole day. In case of flooding collisions were taken into account, resulting in the high percentage of lost messages.... ..."

"... In PAGE 9: ... ing step and each subsequent checkpointing step to be roughly equal in duration. Table4 corroborates this expectation | from columns three and six, we cal- culate that the initial checkpointing of A and b takes approximately 38 seconds on the Sparc-2 apos;s and 30 sec- onds on the SP2, and each checkpoint of x, p, r, w and takes approximately 25 seconds on the Sparc- 2 apos;s and 51 seconds on the SP2. 7 Discussion The results presented in the previous section show that on current NOWs, the performance of this method for fault-tolerant computation is very good.... ..."

"... In PAGE 8: ... Figure 11 is more suitable for loss- tolerant applications to used as a guideline in maximizing the throughput in broadcasting. Table2 and 3 constitutes the nal output of our simula- tions and can be used as an input to adaptive packet type and NBC determination algorithms. 5.... In PAGE 9: ... An important parameter of the current Blueooth broad- casting scheme is the NBC value that is determined by the master of a piconet. We tabulate in Table2 the optimal packet types and NBC values to be used at Bluetooth base- band layer while transferring broadcast application traf c over noisy Blueooth radio channels with various BER val- ues. Based on this information, the baseband layer can adapt to BER variations to achieve a good trade-off between effective throughput and reliability.... ..."