Anonymous Gossip:Improving Multicast Reliability in Mobile Ad-Hoc Networks?Ranveer Chandra Venugopalan Ramasubramanian Kenneth P.Birman
Department of Computer Science
Cornell University,Ithaca,NY14853,USA
{ranveer,ramasv,ken}@https://www.doczj.com/doc/2f13374349.html,
Abstract
In recent years,a number of applications of ad-hoc net-works have been proposed.Many of them are based on the availability of a robust and reliable multicast protocol.In this paper,we address the issue of reliability and propose a scalable method to improve packet delivery of multicast routing protocols and decrease the variation in the number of packets received by different nodes.The proposed pro-tocol works in two phases.In the?rst phase,any suitable protocol is used to multicast a message to the group,while in the second concurrent phase,the gossip protocol tries to recover lost messages.Our proposed gossip protocol is called Anonymous Gossip(AG)since nodes need not know the other group members for gossip to be successful.This is extremely desirable for mobile nodes,that have limited resources,and where the knowledge of group membership is dif?cult to obtain.As a?rst step,anonymous gossip is implemented over MAODV without much overhead and its performance is studied.Simulations show that the packet delivery of MAODV is signi?cantly improved and the varia-tion in number of packets delivered is decreased.
1.Introduction
Ad-hoc networks are mobile networks that operate in the absence of any?xed infrastructure,employing peer-to-peer communication to establish network connectivity.These networks have a wide range of applications such as disaster relief and?eld operations,war front activities,and commu-nication between automobiles on highways.Group com-munication or multicast is a natural requirement for many of these applications and the reliability of the multicast pro-tocol could affect their performance signi?cantly.Ad-hoc networks function under severe constraints such as mobility of nodes,insuf?cient power and memory on mobile devices, and bandwidth restriction of the wireless medium.These ?The authors were supported in part by DARPA/AFRL-IFGA grant F30602-99-1-0532and in part by NSF-CISE grant9703470,with addi-tional support from the AFRL-IFGA Information Assurance Institute,from Microsoft Research and from the Intel Corporation.restrictions make the existing multicast routing protocols such as MAODV very unreliable even in moderately sized networks.This paper discusses our protocol,Anonymous Gossip,a scalable method for providing probabilistic guar-antees to multicast reliability in mobile ad-hoc networks.
2.Multicast Reliability in Ad-Hoc Networks
Several protocols have been designed in recent years to address the issues of multicast routing in ad-hoc net-works.Approaches range from simple ideas such as se-lective?ooding[13]to more complex protocols that main-tain knowledge of the network connectivity or dynamically gather route information.Many multicast protocols ap-proach the problem of multicast in ad-hoc networks by building and maintaining multicast trees or meshes to estab-lish connectivity among group members.MAODV[2]and AMRIS[8]are protocols that maintain multicast trees while ODMRP[10]and MCEDAR[6]are mesh-based.The mesh-based protocol,ODMRP[10]provides better packet deliv-ery than tree-based protocols but pays an extra cost for mesh maintenance.However,these protocols do not attempt to ensure packet delivery and packet loss is a problem during mesh/tree recon?guration,a frequent repair activity.Fur-ther,the number of packets received by different members of a group is highly variable,with some members receiving very few packets while others receive almost all the packets even though the network may not be partitioned at that time.
Multicast reliability in wired networks has received a lot of attention in the past years.Protocols such as SRM[3], RMTP[4]and PGM[5]focus on best effort reliability but are very scalable and easy to implement.These protocols, which are NACK-based(meaning that the receiver has the onus for initiating recovery)depend on the multicast rout-ing tree constructed by the Internet group multicast proto-col,IGMP.In ad-hoc networks,routes changes very rapidly and the methods used by these protocols are consequently not available to us.Bimodal Multicast[1]is another scal-able system,which uses gossip to provide probabilistic re-liability in wired networks.Our premise is that gossip is well matched to the needs of ad-hoc networks because it is a controlled form of?ooding-messages are slowly prop-
agated through the network without congesting the wire-less medium-and is independent of topology.This paper proposes a new method of gossip called anonymous gos-sip(AG),which does not require a group member to have any knowledge of the other group members.
A multicast protocol based on anonymous gossip would proceed in two phases.In the?rst phase,packets are mul-ticast to the group using any unreliable multicast protocol. In the second phase,periodic anonymous gossip takes place in the background and ensures that most of the reachable members receive the packets.This method can be imple-mented on top of any of the tree-based and mesh-based pro-tocols with little or no overhead,and without affecting the scalability of the underlying protocol.In this paper,we dis-cuss our implementation of AG using MAODV as the un-derlying protocol.Simulation results show that using gos-sip over MAODV signi?cantly increases the packet deliv-ery,while the variation in the number of packets received by different nodes decreases.
3.Overview of AODV Multicast
MAODV is a reactive protocol that dynamically creates and maintains a multicast tree for each group.It is an adap-tation of AODV,a unicast routing protocol.Due to con-straints of space,we present in this section a brief overview of only those aspects of MAODV relevant to our implemen-tation.A detailed description of MAODV can be found in [2][11].
Each node running MAODV maintains two routing ta-bles:Route Table(RT)and Multicast Route Table(MRT). The Route Table is used for recording the next hop for routes to other nodes in the network.Each entry in RT con-tains a destination IP address,a destination sequence num-ber,hop count to the destination,IP address of next hop,and the lifetime of this entry.The destination sequence number tracks the freshness of the route to that destination.A source node S trying to send a message to a node B,?rst looks for a route to B in its RT.If a valid route is not found,S broad-casts a route request message called RREQ.A node receiv-ing this RREQ message can unicast a route reply RREP to S if it is the destination node or if it has a fresh enough route to B.Otherwise the node broadcasts the RREQ to its neighbors.The source node S selects the shortest among the freshest routes from the received RREPs and adds the entry in the Route Table.Nodes relaying the RREQs and the RREPS add the reverse and forward route entries into their Route Table respectively.
The Multicast Route Table contains entries for multicast groups of which the node is a router(i.e.,a node in the mul-ticast tree).Each entry in this table contains the multicast group IP address,the group leader IP address,the group sequence number,hop count to the group leader,the next hops,and the lifetime.The next hops are the nodes in the multicast tree to which this node is connected.Each next hop entry has an enabled?ag to indicate a potential but not yet activated entry.The next hop that is closer to the group leader is called the upstream node.
A node S that is not a part of the multicast tree can join the multicast group by broadcasting a RREQ message with the join?ag set.Any node in the multicast tree can respond to a Join RREQ by unicasting an RREP back to S.These RREQs and RREPs are processed similar to unicast routing. In addition,nodes receiving Join RREQs also add entries with enabled?ag false in their MRT.The node S selects a suitable route from the RREPs and sends an activation message called MACT along this route.All nodes receiving the MACT message change the enabled?ag to true in their entries.
Any group member,which is a leaf node in the multi-cast tree can leave the group by sending a MACT message to its upstream node with the prune?ag set.A node re-ceiving a Prune MACT deletes the sender from its next hop table.If it is a non-group member that has now become a leaf node,it leaves the group by sending a Prune MACT to its upstream node.Non-leaf nodes can leave a multicast group but must continue to function as routers in the multi-cast tree.When a link breakage occurs between two nodes U and D of a multicast tree,only the downstream node D attempts to repair this link.This restriction is necessary to prevent formation of loops.D sends an RREQ with an ex-tension containing the hop count to the group leader.Any multicast tree member closer to the group leader than D can reply to this RREQ.In case D receives no replies within a certain time even after a few rebroadcast of the RREQ,the network is assumed to be partitioned and a new group leader is selected in the downstream sub-tree.The details of how the new group leader is selected and how two partitioned trees can rejoin are described in[2][11].
4.Anonymous Gossip Protocol
Gossip as a general technique has been used to solve a number of problems such as network news dissemina-tion(NNTP),replicated data management[14]and failure detection[15].Bimodal multicast[1]uses gossip as a technique to achieve probabilistic reliability of multicast in wired networks.This protocol achieves a bimodal guaran-tee i.e.,all or no delivery with very high probability and partial delivery with very low probability,without sacri?c-ing scalability and stable throughput(low jitter).We also use gossip to address the problem of reliable multicast in mobile ad-hoc networks and provide all or no delivery with very high probability and partial delivery with very low prob-ability.However,we use a different method of gossip to provide the same guarantees.
A gossip based reliable multicast protocol involves two phases.In the?rst phase,any suitable unreliable protocol is used to multicast the message m,to be sent to the group. In the second phase,gossip is used to recover lost mes-
sages from other members of the group that might have re-ceived it.This phase consists of periodically repeated gos-sip rounds in the background as more and more messages are multicast.A single gossip round can potentially recover many lost messages.A single round of gossip consists of the following steps,
1.Node A randomly chooses another member of the
group,say B.
2.A sends B the information about messages it has re-
ceived or not received.
3.B checks to see if it has received any of the messages
listed by A.
4.Then A and B could exchange messages which are not
a part of each other’s message history.
Posing such a broad algorithm in the form of a proto-col has to be done with meticulous care.The constraints of the environment have to be taken into account.As we have mentioned earlier,one of our primary goals is to pro-vide reliability without increasing the message overhead of the already congested mobile wireless environment.The following issues have to be answered keeping in mind the above constraints.
1.How does node A know who the other members of the
group are?
2.How does A select which member to gossip with?
3.What should be the direction of information exchange?
4.How is the message history maintained?
The following subsections describe these design issues in detail.
4.1.Anonymous Gossip
In our opinion,the?rst of the above questions is the most crucial issue.Bimodal multicast requires each participating node to have partial or total knowledge of group member-ship.This usually involves each node sending periodic heart beat messages to other nodes to keep the membership infor-mation current.In wired networks,where the nodes are in a domain sub-domain hierarchy,group membership can be maintained with limited overhead.However,in the wire-less environment maintaining even partial group member-ship is extremely expensive and would signi?cantly reduce the throughput of the network.We propose a method called anonymous gossip(AG)to overcome this problem.AG does not require any member to know the other members of the multicast group.The node attempting to send a gossip mes-sage does not even know the identity of the node with which it will gossip until the other node sends back a gossip reply.
We add a new type of message called gossip message. This message has the following?ve?elds:
/ Router Figure1.Local Anonymous Gossip in MAODV
?Group Address:the address of the multicast group ?Source Address:the address of the node sending the
gossip message
?Lost Buffer:an array of?xed size,that carries se-quence numbers of messages that the source node be-lieves it has lost
?Number Lost:the size of the Lost Array.?Expected Sequence Number:the sequence number of the next message that the source expects.
Each node randomly selects one of its neighbors and sends a gossip message to it.Any node receiving a gossip message randomly selects one of its neighbors(excluding the neigh-bor which sent the message)and propagates the message to it.If the receiving node is itself a member of the multicast group then it randomly decides to either accept the gossip or propagate it.The accepting node then unicasts a gossip reply to the initiator of this gossip request.The reply is de-scribed in section4.4.In a general multicast protocol of an ad-hoc network,the nodes themselves participate as routers. Also,only a subset of these nodes/routers would participate in routing any messages meant for a given multicast group. In such cases,only participating routers are to be consid-ered while propagating the gossip message.For example, in the implementation of this protocol on MAODV,only the routers in the multicast tree participate in propagating the gossip.As have seen in section3,each router maintains a multicast route table which constitutes the nexthops for this router.While propagating the gossip message,one of these nexthops is randomly selected.Propagation along the mul-ticast tree prevents gossip messages from reaching the same node twice.
4.2.Locality of Gossip
Periodic propagation of gossip messages generates con-tinuous traf?c on the network.Choosing nearer members
to gossip with will reduce the network traf?c,however
gossiping with a distant node is extremely important be-cause sometimes message loss could affect an entire local-
ity.Hence we need a scheme that would gossip locally with
a very high probability and with distant nodes occasion-ally.Our AG protocol is augmented to achieve this opti-
mization.Here,we assume some familiarity with MAODV, because the constraint of brevity prevents us from explain-
ing MAODV in detail,and yet our algorithm builds upon
it.We require each node participating as a router in the multicast tree to maintain one additional?eld called near-
est member.In the implementation over MAODV,the mul-
ticast route table is augmented to have this extra?eld asso-ciated with each nexthop entry.The nearest member?eld
associated with a nexthop contains the distance to the near-est group member from this node by taking the link through
this nexthop node.This adds very little overhead to the ex-
isting multicast route table.Whenever a gossip message is received,a nexthop node is chosen so that a nexthop with a
smaller nearest member value is chosen with higher prob-
ability than a nexthop with a larger nearest member value. Thus with higher probability the gossip message is propa-
gated to a closer node than to a distant node.Figure1illus-trates an AODV Multicast Tree,where{A,C,D,H,I,J}
are the group members and{B,E,F,G}are the other par-
ticipating routers.The numbers shown on their edges give the values of the nearest member?eld.As an example,for
the router E,the nearest group member through D is at a
distance1and through F is at a distance3.
Maintaining the validity of the nearest member?eld in each router can be done with limited overhead.This value
needs to be modi?ed when one of the following events oc-curs-a new member joins the group,an existing mem-
ber leaves the group,or the mobility of nodes creates a
topology change.Whenever a new member joins the group or an existing member leaves the group,the routers near-
est to this node realize the event?rst.For example in
MAODV a member sends a MACT message to establish itself while joining(see section3for details).A mem-
ber leaving the group sends a prune message to its nearest routers.The nearest router adds this new nexthop to its mul-
ticast route table with value of nearest member?eld set to
one or deletes this entry from its list of nexthops.Then for each of the nexthops present in its table,it recomputes the
value of nearest member and sends this modi?ed value to
that nexthop.For example,if{B,C,E}are the next hop entries for node D,and{b,c,e}are the associated near-
est member values for these nexthop entries,then D sends 1+min(c,e)to B,1+min(b,e)to C and1+min(b,c) to E.This value needs to be sent only if it is different from
its previous value.Also,this value can be piggy backed on any other messages bound for the same router.When
a router receives this new value,it updates this entry in its table.It then sends a modify message to its nexthops in the same way as described above.Since we are propagat-ing the minimum of a set of values,any change in topology or group membership will only affect the routers in its lo-cality.These modify messages will not propagate far and hence will not increase congestion signi?cantly.
4.3.Cached Gossip
AG is done only using the routers participating in the multicast tree.If these routers are not well distributed in the topology of the network,there is a possibility that a gos-sip message reaches a node along a route longer than the shortest existing route between these two nodes.For ex-ample,when using AG over MAODV,the gossip message propagates only along the multicast tree,although other, potentially shorter,routes may exist between the gossiping members(note that the return path is unaffected by this phe-nomenon,because gossip replies are unicast).It is ef?cient to use the unicast routes to gossip with those nodes whose membership is already known.Further,these unicast routes may be available for gossip even when the multicast tree is being repaired.
We incorporate this by introducing a member cache in all the member nodes of each multicast group.The mem-ber cache is a bounded buffer containing entries which are3-tuples(node addr,numhops,last gossip).The node addr?eld contains the address of a group member, the numhops?eld contains the shortest distance between the nodes and the last gossip contains the time at which last gossip occured between these two nodes.This information itself is collected at no extra cost.The member cache table is updated each time a message is received from a group member.This message could be a data packet meant for this group,a gossip reply,or any other maintenance packet. For example,in MAODV,each route request(RREQ)for a group join generates replies from a number of nodes,many of which are members of the group(see[11][2]for details). Whenever such a message is received,we add the member information to the member cache.If this table is full and a new member has to be added,a member with a greater numhops is deleted from this table.If there are no members with greater numhops then the member with most recent last gossip is replaced with the new entry.This is done to avoid frequent gossips with the same members.
In each gossip round,the node chooses to do anonymous gossip with probability p anon.If AG is chosen,then a gos-sip message is constructed with the address of this node as the initiator and this message is propagated as described in section4.1and4.2.If cached gossip is chosen,then a mem-ber is selected randomly from the member cache and a gos-sip message is unicast to this member.When a gossip reply is received from a member,the member information is up-dated in the member cache.If this member already exists in this table,then the numhops and last gossip?elds are
updated.Otherwise,the member is added to the table as described above.
4.4.Push vs Pull
The importance of direction of information exchange is explained in[14].Our protocol implements a pull mode of message exchange that may be described as follows.Each node maintains a table called lost table,for every multicast group that it belongs to.This contains the sequence num-bers of all the messages this node believes itself to be lack-ing.An entry in this table will be made whenever a mes-sage is received with a sequence number greater than the ex-pected sequence number.Note that the sequence number is a2tuple including the sender address and a sequence num-ber,because different senders send messages to the mul-ticast group.Each node also maintains a bounded FIFO buffer,called history table containing the most recent mes-sages received.The most recent entries of the lost table are placed in a lost buffer.Whenever a node prepares a gos-sip message,the lost buffer and a list of expected sequence numbers is added to the gossip message.When a node re-ceives a gossip message,it compares the lost buffer and the expected sequence number list to see if its history table has a copy of any message sought by the gossip initiator.It then unicasts any message found back to the gossip initiator as the gossip reply.
5.Simulation and Performance
We simulated our protocol on GloMoSim[17],which uses a parallel,event-driven simulation language called PARSEC[18].MAODV was implemented as described in version5of the IETF draft[11]and was adapted to imple-ment our gossip protocol.We understand that more recent drafts of MAODV are available,but we believe that the ex-tensions would not impact performance for the scenarios investigated.
5.1.Simulation Environment
A?xed area of200m×200m was used in the simulation, and all the nodes were initially placed randomly within this area.The simulations were performed using the Random-Waypoint scheme to model the mobility of the nodes in the network.In this model,each node has a prede?ned mini-mum and maximum speed.It then travels towards a ran-dom spot at a speed chosen randomly from this interval. After reaching the destination,it rests for a period chosen from a uniform distribution between0and80seconds,be-fore continuing this mobility pattern.Each simulation was set to run for10minutes.The MAC layer protocol used was IEEE802.11and the bandwidth of the wireless medium was assumed to be2Mbps.
The network consisted of a single multicast group with one-third of the nodes being group members.All the nodes joined the group at the beginning of the simulation and re-mained in it throughout the run period.One of the members was selected as the source of data packets.The source node generated packets of length64bytes at regular intervals of 200ms with the destination as the multicast group.In the 600second simulation time,the source node continued to send packets starting at120seconds and ending at560sec-onds.The initial delay of120sec was allowed for MAODV to build the initial multicast tree.The source node generated 2201data packets during the period of interest.
Every group member sends one gossip message per sec-ond.Each gossip message could request for at most10lost messages.The size of the membership cache was also set to10.Each member could keep track of at most200lost messages,in the lost table,and remember the most recent 100messages received in the history table.For MAODV, the hello interval was set to600ms,the allowed hello loss to4,and the group hello interval to5sec.
We studied the performance of our protocol while vary-ing three parameters:the transmission range,the maximum speed and the number of nodes.The transmission range is the distance from a node within which another node can re-ceive packets from it.All the nodes are assumed to have the same transmission range.The minimum speed was set to0for all the runs.We varied each parameter over a wide range,for different?xed values of the other two parame-ters.For each set of parameters,we measured the number of packets received by each node in the multicast group. Each simulation was carried out10times with different val-ues for the random-seed.
The performance graphs measure the number of packets received by the group members during each run.Each data point in the graph corresponds to the average of the number of packets received by each group member.The range of measured data values obtained for the full set of receivers is shown in the form of error bars,while a line connects the average across receivers.Notice that reception rates can vary widely for different receivers because of differences in network connectivity.Perfect reception rates would be graphed as2201:the number of packets multicast by the source.
5.2.Packet Delivery vs.Transmission Range
The transmission range was varied from45m to85m in steps of5m,keeping the total number of nodes constant at 40.This experiment was carried out for6values of the maximum speed.Fig.2and Fig.3show the variation of packet delivery for maximum speeds0.2m/sec and2m/sec respectively for both MAODV and our gossip protocol.
It is interesting to note that the gossip protocol consis-tently performs better than the underlying multicast pro-tocol.Not only is the average packet delivery higher but the variation in the number of packets received by different group members is also signi?cantly lower.As the transmis-sion range increases,the connectivity of the network im-proves leading to fewer link failures.Further,the length
Figure 2.Packet Delivery vs Transmission Range when the max speed=0.2m/sec
Figure 3.Packet Delivery vs Transmission Range when the max speed=2m/sec
of the network routes becomes shorter,decreasing the route discovery and setup time.The effect of this is re?ected by the steady improvement in packet delivery of both gossip and MAODV with increase in the transmission range.How-ever,this runs a risk of increasing the congestion in the net-work,which could affect the performance of the protocols.
5.3.Packet Delivery vs.Maximum Speed
The maximum speed of the nodes was varied in two phases,from0.1m/sec to1m/sec in steps of0.1m/sec and from1m/sec to10m/sec in steps of1m/sec.The transmis-sion range was set constant at75m and the number of nodes was?xed at40.Fig.4and Fig.5show variation of packet delivery with speed for both the phases.These graphs con-tinue to illustrate that our protocol achieves better packet delivery with a decreased variation in the number of pack-ets received by the group members.At very low values of maximum speed,upto0.3m/sec,our protocol gives near 100%packet delivery.
At higher speeds,the data delivery
Figure4.Packet Delivery vs Maximum Speed when the transmission range is
75m
Figure5.Packet Delivery vs Maximum Speed when the transmission range is75m
varies from90%to80%.We can see a gradual decrease in data delivery as the maximum speed increases in this range. Increase in the maximum speed causes the nodes to move faster and more frequently in the random waypoint model. This causes an increase in the frequency of link breakages, affecting the overall performance of the system.
5.4.Packet Delivery vs.Number of Nodes
The number of nodes was varied from40to100.In one experiment,the transmission range was adjusted in such a way that the average number of neighbors of a node re-mained approximately the same.As the number of nodes in the network increases,the routing distance between the nodes goes up,and hence the frequency of link failures in the network also increases.Fig.6shows the variation of packet delivery with the increase in the number of nodes. As the number of nodes increases,the packet delivery rate tends to decrease gradually because of the above mentioned reasons.In another experiment,the transmission range was
Figure6.Packet Delivery vs Number of Nodes keeping the average number of neighbors for
a node approximately constant
kept constant at55m and the number of nodes was varied from40to100.Fig.7shows the results of this experiment. As the number of nodes is increased,the connectivity of the network increases improving packet delivery.However, the traf?c in the network also increases and beyond a point this leads to congestion decreasing the number of packets delivered.
5.5.Performance Analysis
The simulation results show that AG can improve the re-liability of multicast routing protocols without the use of ac-knowledgements,and without adding signi?cant overhead to them.The overhead analysis of gossip is presented in Fig.
8.Goodput is de?ned as the percentage of non-duplicate messages received through gossip replies to the total num-ber of messages received through gossip replies.In other words goodput gives a measure of the redundant traf?c-more the goodput,more the number of useful messages car-ried by gossip replies and lesser is the redundancy.In our simulations goodput is measured at different group mem-bers for two values each of transmission range and maxi-mum speed.Fig.8shows that the goodput is close to100 percent for most of the cases under which the experiments were run.It implies that nearly all the gossip replies were useful and very few duplicates were received.This can be attributed to the parameters that were chosen for gossip.For a different messaging rate these parameters would have to be adjusted to give a comparable goodput.In general the gossip rate should be tuned so that the network does not get congested and the goodput is nearly100percent.
AG works even when the multicast tree is broken,re-covering packets dropped during the maintenance of the
Figure7.Packet Delivery vs Number of Nodes when the transmission range is constant at 55m and maximum speed is0.2m/sec
tree.The effectiveness of anonymous gossip depends on the values chosen for the size of the history table and the lost table,besides the gossip interval.We are currently doing more experiments to study the performance of AG under different parameters.Implementing anonymous gos-sip with other multicast protocols,such as ODMRP[10]and AMRIS[8],could also be done in a similar manner without adding extra complexity.We believe that AG would be able to improve the reliability of these protocols as well.
6.Related Work
Despite the importance of reliable multicast in mobile ad-hoc applications,not much work has been done on it.In [9],Pagani et al propose a protocol that guarantees validity, agreement,integrity and termination for multicast in such environments.The protocol is characterized in two phases: the scattering phase,in which a message is diffused from the source to all its destinations,and the gathering phase, in which the source collects all the acknowledgements mes-sages from the destinations.Most of the communication is done on a forwarding tree,that is an on demand source-centered routing tree,with the lifetime of at most the time needed to scatter and gather a single message.In case,the tree links break,no effort is made to repair it,and?ood-ing is used for further communication.For reliability,a message is buffered at each of the clusterheads along the tree,as long as its ack from the downstream nodes is not https://www.doczj.com/doc/2f13374349.html,ing this mechanism,the protocol provides strong guarantees for message delivery as long as the conditions of eventual subsidence,liveness and clusterhead stability is satis?ed.While this protocol provides these guarantees in a slow moving network where the failures are rare,the per-
Figure8.Goodput at different group members
for2values each of transmission range and maximum speed
formance decreases dramatically with the increasing size of the network.There is also degradation in performance with increase in mobility.The reasons for the poor perfor-mance can be attributed to two main aspects of the protocol. Firstly,since buffer size,in practice,is bounded,a situation could arise when old messages are still stored in most of the node buffers.This gives rise to a situation where newer messages cannot be accommodated since older messages have to be stored to provide the strong guarantees.The sec-ond reason for decrease in performance is the use of ack messages.This proves to be very expensive in wireless net-works where the physical layer is bandwidth constrained.
Another approach to achieve reliability in ad-hoc net-works is proposed in[13].This protocol guarantees best effort delivery using an adaptive?ooding scheme called hyper-?ooding.The nodes do not keep any state regarding the multicast routes.Whenever a sender has to multicast a message,it broadcasts the message to all its neighbors who in turn re-broadcast the packet.Some optimizations are made to prevent packets from looping forever in the network.The advantage of this protocol is the reliability it provides for high-speed networks.On the other hand this protocol is extremely expensive since it generates a large number of messages,and may easily congest the network. Especially when the multicast group is sparsely distributed in a dense network or when all the nodes are not moving at high speeds,the protocol decreases the throughput consid-erably.
7.Concluding Remarks and Future Work
Many of the exciting applications for wireless ad-hoc networks will require reliable multicast although the degree of reliability demanded by them may vary.Few applica-tions might demand special properties from the protocol, many may be satis?ed with best effort delivery and others could be satis?ed with an unreliable multicast as long as most of the messages reach the destination.While[9]de-scribes a protocol for the?rst case,we do not address that problem here.At the same time we would like to emphasize that stronger reliability guarantees can be obtained by devel-oping protocols that function over the underlying multicast and our gossip protocol.Work is being done on the Sp-inglass project[16]in Cornell University,where protocols for leader election,consensus and other advanced proper-ties are being developed over bimodal multicast.As a fu-ture work,we would like to develop ef?cient algorithms to do the same over probabilistic multicast protocols in mobile ad-hoc networks.
In this work,we have not tried to improve upon the ef?-ciency and scalability of MAODV itself.Our protocol could be used with any existing multicast protocol,without signif-icantly changing the existing protocol.AG inherits limita-tions from the underlying protocol;for example,in the work described here,AG would not work if the links were unidi-rectional.On the other hand AG introduces little overhead beyond that associated with the underlying mechanisms.In the future,with much better and ef?cient multicast proto-cols coming up,we would like to improve upon their relia-bility by using the gossip protocol.Our work with MAODV is the?rst effort in improving reliability of the existing mul-ticast protocols.It is possible to get improvements by using anonymous gossip on the other protocols proposed in[6], [10],[8]and[7].
Acknowledgements
We are extremely grateful to Indranil Gupta and Robert Van Renesse for the useful discussions and Chaitanya Swamy and Sheela Tiwary for reviewing the paper. References
[1]Kenneth P.Birman,Mark Hayden,Oznur Ozkasap,
Zhen Xiao,Mihai Budiu and Yaron Minsky.“Bimodal Multicast”.ACM Transactions on Computer Systems, V ol.17,Issue2,May1999,41-88.
[2]Elizabeth M.Royer and Charles E.Perkins.“Multicast
Operation of the Ad-hoc On-Demand Distance Vec-tor Routing Protocol”.Proceedings of the?fth annual ACM/IEEE international conference on Mobile com-puting and Networking(MOBICOM).Aug.1999,207-218.
[3]Sally Floyd,Van Jacobson,Steve McCanne,Ching-
Gung Liu and Lixia Zhang.“A Reliable Multicast Framework for Light Weight Sessions and Application Level Framing”.In Proceedings of the’95Symposium on Communication Architectures and Protocols(SIG-COMM).ACM.August1995,Cambridge MA.
[4]Sanjoy Paul,K.K.Sabnani,J.C.Lin,S.Bhattacharya.
“Reliable Multicast Transport Protocol(RMTP)”.IEEE
Journal on Selected Areas in Communications,special issue on Network Support for Multipoint Communica-tion,April97,V ol15,No.3.
[5]T.Speakman and D.Farinacci and S.Lin and A.
Tweedly.“Pretty Good Multicast”.Internet draft, ftp://https://www.doczj.com/doc/2f13374349.html,/internet-drafts/draft-speakman-
pgm-spec-00.txt
[6]Prasun Sinha,Raghupathy Sivakumar,Vaduvur Bhar-
gavan.“MCEDAR:Multicast Core-Extraction Dis-tributed Ad-Hoc Routing”.Proceedings of IEEE Wire-less Communications and Networking Conference, September1999.
[7]L.Ji and M.S.Corson.“A Lightweight Adaptive Mul-
ticast Algorithm”.Proceedings of IEEE GLOCECOM, Sydney,Australia,December1998,1036-1042.
[8]Chun-Wei Wu,Y.C.Tay.“AMRIS:A Multicast Pro-
tocol for Ad-Hoc Wireless Networks”.Proceedings of MILCOMM’99,Nov.1999.
[9]Elena Pagani and Gian PaoloRossi.“Reliable broadcast
in mobile multihop packet networks”.Proceedings of the third annual ACM/IEEE international conference on Mobile computing and networking(MOBICOM), September26-30,1997,34-42.
[10]S.-J.Lee,M.Gerla,and C.-C.Chiang.“On-Demand
Multicast Routing Protocol(ODMRP)”Proceedings of IEEE WCNC’99,Sep.1999.
[11]C.Perkins, E.Royer,S.Das“Ad-Hoc On De-
mand Distance Vector(AODV)Routing”.IETF Internet Draft,draft-ietf-manet-aodv-05.txt,March14,2000. [12]Elizabeth M.Royer and Charles E.Perkins.“Ad-hoc
On-Demand Distance Vector Routing”.Proceedings of the2nd IEEE Workshop on Mobile Computing Systems and Applications,Feb.1999,90-100.
[13]Christopher Ho,Katia Obraczka,Gene Tsudik and
Kumar Viswanath.“Flooding for Reliable Multicast in Multi-Hop Ad-Hoc Networks”.MobiCom Workshop on Discrete Algorithms and Methods for Mobility,Aug 1999.
[14]Alan Demers,Dan Greene,Carl Hauser,Wes Irish
and John Larson.“Epidemic algorithms for replicated database maintenance”.Proceedings of the Sixth An-nual ACM Symposium on Principles of distributed com-puting,1987,1-12.
[15]Robbert van Renesse,Yaron Minsky,and Mark Hay-
den.“A Gossip-Based Failure Detection Service”.Proc.
of Middleware,1998.[16]Cornell University.“Spinglass:Adaptive Probabilistic
Tools for Advanced Networks.”
[17]Xiang Zeng,Rajive Bagrodia,and Mario Gerla.“Glo-
MoSim:A Library for Parallel Simulation of Large-scale Wireless Networks”.Proceedings of the12th workshop on Parallel and distributed simulation,May 26-29,1998,Banff Canada,154-161.
[18]R.Bagrodia,R.Meyer,M.Takai,Y.Chen,X.Zeng,
J.Martin,B.Park,and H.Song.“Parsec:A Parallel Simulation Environment for Complex Systems”.IEEE Computer,Vol.31(10),October1998,pp.77-85
美剧口语：《绯闻女孩》第一季地道口语（1） Lily: Serena you’ve been gone, doing who knows what with god knows who. Serena你一直都不在这里,在鬼知道什么是哪儿的地方做鬼知道是什么的事儿. * 很可爱的说法吧。 这里有同学提出疑问，某迷核对过了，是原文，you've been gone里的这个gone个人理解是做形容词（额，就是那个类型的了。不懂语法，差不多就是那个意思吧。。），就像avril有首歌叫when you're gone，同理。恩。 Chuck: Any interest in fresh air? 想出去走走透透气吗？ * 好了，邀请别人终于可以不用“do you want to/would you wanna……”了。ps，这里C的意思是出去抽烟。。 Gossip: Game on. 游戏开始了。
* on的用法如果全部都掌握，那么很多句子的表达都会更简洁了。比如“穿鞋”，就不用说“put your shoes on”而可以简单地说，“shoes on” 了。 Eric: No offense. Blair: None taken. ——没有冒犯的意思。 ——我不介意。 * 这是很经典的问答。好好记住了。。 Chuck: Move, plz. 请让一下。 * 如果是某迷，之前只知道，“excuse me”和“outta my way”两个说法。目前觉得这个很可爱。ps说这个可爱的原因是觉得它体现了C的风格，贵族式的傲慢。 Blair: Done and done. 成交！ * 成交还有其他的说法，比如单用“Deal”。不过觉得大部份人应该都知道吧，下次可以用这个，比较有新鲜感了。。嗯。。 Blair：Nighty-night. 晚安。 * 可爱的说法吧~女孩子好好记下~。
《断舍离》经典语录,让自己活得更开心,没 事常看看
《断舍离》经典语录，让自己活得更开 心，没事常看看 导读: 《断舍离》经典语录，让自己活得更开心，没事常看看 1、断，断绝不需要的东西。舍，舍去多余的废物。离，脱离对物品的执着。现在对自己来说不需要的就尽管放手。 2、不管东西有多贵，有多稀有，能够按照自己是否需要来判断的人才够强大。能够放开执念，人才能更有自信。 3、断舍离的主角并不是物品，而是自己，而时间轴永远都是现在。选择物品的窍门，不是“能不能用”，而是“我要不要用”，这一点必须铭刻在心。 4、从加法生活转向减法生活很重要，并不是心灵改变了行动，而是行动带来了心灵的变化。可以说，断舍离就是一种动禅。 5、断舍离，就是透过整理物品了解自己，整理内心的混沌，让人生舒适的行动技术。换句话说，就是利用收拾家里的杂物来整理内心的废物，让人生转而开心的方法。 6、人类最大的罪是不快活，让人变得快活是让一切变好的先决条件。我们很容易让自己随着别人的心情起伏波动，让自己陷入别人的引力圈。其实不应该这样，应该把不快活的人拉到自己的心情快活引力圈里来。
7、断舍离在行为上要先学会“舍”，也就是把不需要的东西全部扔掉。舍的秘诀，就是完全地以自己为中心，并且以当下为时间轴。因此， 1、在考虑物品是否应该被留下时，思考的主语是“我”，而不是物品。 2、将与物品的关系比作人际关系，选择对当下的我来说最必要的朋友。更高级别是只选择真正必需、而且自己又喜欢的东西的阶段。扫除：分为收拾，整理，以及表现为扫、擦、刷的打扫。 8、三种扔不掉东西的人： 1、逃避现实型——不愿待在家里2、执着过去型——对过去幸福时光的留恋 3、担忧未来型——致力于投资未来的不安因素对现在的界定因人而异扔不掉是自己把感情移到物品上面，并因此充分了自己扔不掉=不想扔的机制破烂分三类： 1、不用的东西——会念咒的束缚人的淤泥 2、还在用的东西——一滩混乱的淤泥 3、充满回忆的东西——散发出强大的气场堆积的破烂代表着良心不安的聚集。 9、这些在居住环境里放了好几个月，甚至放了好几年的东西，只不过因为不是生鲜食品所以才没有烂掉。但是如果从机能上来说，它们早已经腐烂了。置身于这样的环境中，就几乎等于是暂住在一个垃圾暂放室。主语到底是“我”还是“物品”？请养成经常如此自问的习惯吧。 10、断舍离是生活的减法哲学，减去多余的物品，认清自我，磨砺感知的本能；俯瞰力是心灵的加法哲学，了解深层需求，
柳传志，1944年4月29日出生于江苏镇江，曾任联想控股有限公司总裁、董事局主席，2011年11月2日卸任。现任联想集团有限公司董事局名誉主席，联想集团高级顾问。企业家、投资家、泰山会成员、全球CEO发展大会联合主席。 1、人生在世，要受许多委屈。而一个人越是成功，所遭受的委屈也越多。要使自己的生命获得极值，就不能太在乎委屈，不能让它们揪紧你的心灵。要学会一笑置之，要学会超然待之，要学会转化势能。智者懂得隐忍，原谅周围那些人，让我们在宽容中壮大。 2、只要老倪说的都对，老倪是有效数字1，别的科技人员都是0，这些0只有跟着倪光南干才能出成绩。 3、要有理想，但是不要理想化! 4、恶劣的环境下，搭建一个温室，让幼苗无所顾忌地成长。成长过程中，适合我的留下，不适合的扔掉。 5、第一年根本谈不上战略。 6、搭班子，定战略，带队伍。 7、看油画的时候，退到更远的距离，才能看明白。离得很近，黑和白是什么意思都分不清。退得远点，就能明白黑是为了衬托白，再远点，才能知道整幅画的意思。打这个比喻是为了时时提醒我们牢记目标，不至于做着做着就做糊涂了。 8、你(杨元庆)知道我的大鸡和小鸡的理论。你真的只有把自己锻炼成火鸡那么大，小鸡才肯承认你比他大。当你真像鸵鸟那么大时，小鸡才会心服。只有赢得这种心服，才具备了在同代人中做核心的条件。 9、很难想象，一个企业既做电脑，又做房地产，还做投资，然后用同一套体系去进行人事建设和激励。让这样的企业与高度专业化的企业竞争，想取胜是非常困难的多元化非常忌讳采用企业事业部的方式由企业总部给各事业部制定战略，进行激励。而采取子公司的形式就不同。子公司是独立法人，可以有独立的战略体系、执行体系和文化建设体系。 10、(联想并购IBM PC事业部)到目前为止，取得这样的成果，我总结了一条经验：就是预先要把事情想清楚，把战略目的、步骤，尤其是出了问题如何应对，一步步一层层都想清楚;要有系统地想，这不是一个人或者董事长来想，而是有一个组织来考虑。当然，尽管不可能都想得和实际中完全一样，那么意外发生时要很快知道问题所在，情况就很好处理了。 11、(企业家的主要失败原因)一是没有把目标想清楚。有很多企业家做企业，做着做着就想向政治靠拢了，有心参与政治，结果却把企业做死了;二是一些企业家拿了不属于自己的东西，侵犯了股东的，甚至是国家的利益;三是把长跑当成了短跑。企业成长犹如跑步，倘若
Lesson 1 Excuse me! 对不起！ Lesson 2 Is this your…这是你的……吗？ 一、教学目的 1．理解主系表结构的陈述、一般疑问式 2．初步掌握询问“是否”的方法 二、教学重点 1．主系表结构 2．一般疑问句 3．代词小引 三、教学难点 1．人称代词和物主代词的区分运用 2．主系表结构及其一般疑问句 四、教学标准 1．识别并简单运用主系表结构 2．掌握一般疑问句及肯定回答 五、教学内容 1．语法点： 1)代词人称代词和物主代词 2)Be 动词 3)一般疑问句 4)主系表结构this为主语，名词做表语 5)一般疑问句以及它的肯定回答
2．语言点： 1)打扰他人 2)表达谢意 3．语音：初步知识，字母、音标介绍六、扩展练习 1．打扰他人的表达 2．表达谢意
Lesson 3 Sorry, sir. 对不起，先生。Lesson 4 Is this your…这是你的……吗？ 一、教学目的 1.理解并运用主系表结构的陈述、一般疑问式 2.掌握询问“是否”的方法 二、教学重点 1.主系表结构 2.一般疑问句 3.代词形容词性物主代词 三、教学难点 1.人称代词和物主代词的区分运用 2.主系表结构及其一般疑问句 四、教学标准 1.识别并简单运用主系表结构 2.掌握一般疑问句及否定回答 五、教学内容 1.语法点： 1)否定陈述句 2)Be动词的否定形式 3)代词的功能 4)一般疑问句及否回答 5)形容词性物主代词 2.语言点：
1)询问“是否” 2)表达歉意 3.语音：字母、音标介绍 六、扩展练习 1.询问“是否” 2.向别人道歉
第二集 ur one and only source into the scandalous lives of manhattan's elite. top story on my home page-- serena van der woodsen, everybody's favorite "it" girl, has just returned from a mysterious absence. you'll never believe what's on "gossip girl." someone saw serena getting off a train at grand central. and everyone is talking. blair,it's serena. serena? and b.'s boyfriend nate-- rumor has it-- he's always had a thing for serena. but you're back now. i didn't come back for you. look, blair's my best friend, and you're her boyfriend, and she loves you. why did she leave? why did she return? must be a lot of rumors why you're back. yeah, but none of them mention you. he tries to take his own life, and you're worried it's gonna cost you, mom? you have no idea what it's been like. the best friend and the boyfriend-- that's pretty classy, s. and nate's friend chuck won't let serena forget about her past. i'm trying to change. i liked you better before. and then there's dan, the outsider. looks like his childhood crush has returned. she doesn't know me. nobody knows me. it's cool. no. no, no, no, no, no. uh, serena, um, do you know this young man? oh, from last night. you remember me? you'd really go out with some guy you don't know? well, he can't be worse than the guys i do know. guess whose dad is cool. it's a trick question. yeah, 'cause it can't be ours. "top ten forgotten bands of the '90s." yeah, check out who's number nine. he's very proud. hey! hey, way to be forgotten. why is my daughter going to one of your concerts? well, our kids were bound to meet. it's a small island. are you sure it's not some ploy, your using my daughter to get to me now that your wife left you? serena and her mystery man made a surprise appearance at blair's kiss on the lips party. serena's here? what is she doing here? she wasn't invited. but serena wasn't the only one who made an impression... who's the newbie?
《吸引力法则》经典语录 2014-05-23汇聚智慧 1.你生命中所发生的一切，都是你吸引来的。 2.我想什么，我就能得到什么！ 3.宇宙中最强有力的法则就是吸引力法则。 4.同类相吸。 5.思想变成实物。
6.改变了思想，就改变了命运！ 7．所有美好的思想都是强有力的，所有负面的思想都是脆弱无力。 8．主宰心灵的方法：静心！静心的力量极为强大！ 9．我是自己思想的主人！我能完全立刻控制我的思想！ 10．用持续的思想召唤！事情的起因永远都是思想！ 11．感觉要好！增加对事物的渴望和感觉。 12．你拥有改变一切的力量！ 13．思想与爱的融合，形成了吸引力法则不可抗拒的力量 14.爱是宇宙中最伟大的力量，爱的感觉是最高的频率，如果能爱所有的事物和人，你的生命必将转变！ 15.充满爱的思想----天下无敌！ 16.生命中所发生的一切，都与你的目标相关联！ 17.一旦你真正主宰你的思想和感觉，你就是你自己现实的创造者！ 18.这个精彩非凡的宇宙能带给我们所有美好的事物，并暗中协助我们成就每一件事！ 19．宇宙的声音：“你的愿望，就是我的指令。 20．要求-----相信-----接收！ 21．养成要求的习惯------决不动摇的相信-----开心的接收！ 22．“受到启发的行动”毫不费力，在做接收的动作，感觉棒极了！ 23．宇宙喜欢快速行动，不要拖延，不要猜测，立即行动！ 24．心想事成，有求必应。
25．强效的方法：感恩！感恩能转变你的能量，改变你的想法！ 26．感恩的力量胜过其他一切。 27．为你已经拥有和想要拥有的事物感 28．感恩什么就会得到更多什么，感恩越多，得到越多！ 29．感恩感觉视觉化看得见。 30．“我说不出这股力量是什么，我知道它存在。” 31．“想象力就是一切，它是生命将发生之事的预览。” 32．“成功来自于内在，而非外在。” 33．“我是吸引钱的磁铁，我爱钱，钱也爱我，我每天都在接收钱 34．“给予”是把更多金钱带进你生命最强效的方法。 35．用爱和尊重对待自己和别人，才能赢得别人的爱和尊重！ 36．赢得人际关系：欣赏别人！不要抱怨！ 37．焦点集中在爱上，就会有更多爱和喜悦回到你身上！ 38．所有的压力都是由一个负面的思想开始的！ 39．爱和感恩可以解除所有的负面性，并消除任何疾病！ 40．笑，是最佳的良药！笑的时候可以释放所有的消极和疾病！41．每个不愉快的思想，都是放进身体里的坏东西。 42．任何事物，专注它，就是在创造它！ 43．富足的方法：想着富足，看着富足，感觉富足，相信富足！
朗读者经典语录 《朗读者》第一期【主题词：遇见】 1、只要你是个正直的孩子，不管你从事什么行业，你都是我的好孩子。——柳传志父亲对柳传志说的话 2、人生的不可测，在任何时候，都要抱着一份希望。——董卿 3、从某种意义上来看，世间一切，都是遇见。就像，冷遇见暖，就有了雨；春遇见冬，有了岁月；天遇见地，有了永恒；人遇见人，有了生命。——董卿 4、如果美丽只是简单的用来做展示，也许它的意义不是那么大，但如果美丽，能够转化成一种能力，去帮助更多的人，甚至去让自己变得更好，那它就是很有价值很有意义的了。——董卿 5、愿你慢慢长大，愿你有好运，如果没有，希望你在不幸中学会慈悲；愿你被很多人爱，如果没有，希望你在寂寞中学会宽容。——张梓琳朗读刘瑜的《愿你慢慢长大》 6、生命哪，并不是你活了多少日子，而是你记住了多少日子，要使你过的每一天，都值得回忆。——许渊冲 《朗读者》第二期【主题词：陪伴】 1、“妈妈，你到底为什么爱我？”母亲放下针线，用她的面额，抵住我的前额，温柔的，不迟疑地说：“不为什么，只因你是我的女儿！”——杨乃斌朗读冰心的散文《不为什么》 2、这个世界上绝了哪种生命形式，都会导致地球毁灭。狮子和蚂蚁一样伟大，小草和人类一样重要。——郑渊洁 3、我要开花，是为了完成作为一株花的庄严生命，不管你们怎么看我，我都要开花！——林清玄《百合花开》 4、要是没有兔子和鹧鸪，一个田野还成什么田野呢？他们是最简单的土生土长的动物。——梭罗《瓦尔登湖》 《朗读者》第三期【主题词：选择】
1、有人说“不疯魔，不成活”，我很享受这种若隐若现当中左右徘徊，这是我表演的语境，也是我的选择。——王千源 2、你用最不敬业的精神去演戏，结果世界上最单纯的观众，用这么炙热的心情去观看，真不次于在犯罪。——王千源 3、我热爱这种浑然天成的泥泞，泥泞诞生了跋涉者，它给忍辱负重者以光明和力量，给苦难者以和平和勇气。一个伟大的民族需要泥泞的磨砺和锻炼，它会使人的脊梁永远不弯，使人在艰难的跋涉中懂得土地的可爱、博大和不可丧失，懂得祖国之于人的真正含义：当我们爱脚下的泥泞时，说明我们已经拥抱了一种精神。——秦？飞等朗读迟子建的《泥泞》 4、可以说脚踏大地，泥泞，我相信一定会磨砺他们的青春，让他们的人生焕发出与众不同的光华，这光华也必将与我们的时代交相辉映。——董卿 5、人的一生，总是在寻找一种平衡，忠贞的人，永远会得到忠贞；勇敢的人，最后也是用勇敢来结束。——麦家 6、我慢慢去想奶奶讲的那个神话，我慢慢相信，每一个活过的人，都能给后人的路途上添些光亮，也许是一颗巨星，也许是一把火炬，也许只是一支含泪的蜡烛。——《奶奶的星星》 17、一个人最重要的不是拥有多少财富，和拥有多大的房子，最重要的是拥有深厚的学识和崇高的思想道德。——康震点评《陋室铭》 8、有人说，我们这个时代不缺机会，所以也势必会让每个人面临很多的选择，那么，是遵从自己的内心，还是随波逐流；是直面挑战还是落荒而逃；是选择喧嚣一时的功利，还是恒久平静的善良；无论如何，希望每一个人，都能做出一个在日后回想不让自己后悔的选择。——董卿 《朗读者》第四期【主题词：礼物】
《断舍离》经典语录，让自己活得更开心， 没事常看看 导读: 《断舍离》经典语录，让自己活得更开心，没事常看看 1、断，断绝不需要的东西。舍，舍去多余的废物。离，脱离对物品的执着。现在对自己来说不需要的就尽管放手。 2、不管东西有多贵，有多稀有，能够按照自己是否需要来判断的人才够强大。能够放开执念，人才能更有自信。 3、断舍离的主角并不是物品，而是自己，而时间轴永远都是现在。选择物品的窍门，不是“能不能用”，而是“我要不要用”，这一点必须铭刻在心。 4、从加法生活转向减法生活很重要，并不是心灵改变了行动，而是行动带来了心灵的变化。可以说，断舍离就是一种动禅。 5、断舍离，就是透过整理物品了解自己，整理内心的混沌，让人生舒适的行动技术。换句话说，就是利用收拾家里的杂物来整理内心的废物，让人生转而开心的方法。 6、人类最大的罪是不快活，让人变得快活是让一切变好的先决条件。我们很容易让自己随着别人的心情起伏波动，让自己陷入别人的引力圈。其实不应该这样，应该把不快活的人拉到自己的心情快活引力圈里来。 7、断舍离在行为上要先学会“舍”，也就是把不需要的东西全
部扔掉。舍的秘诀，就是完全地以自己为中心，并且以当下为时间轴。因此，1、在考虑物品是否应该被留下时，思考的主语是“我”，而不是物品。2、将与物品的关系比作人际关系，选择对当下的我来说最必要的朋友。更高级别是只选择真正必需、而且自己又喜欢的东西的阶段。扫除：分为收拾，整理，以及表现为扫、擦、刷的打扫。 8、三种扔不掉东西的人：1、逃避现实型——不愿待在家里2、执着过去型——对过去幸福时光的留恋3、担忧未来型——致力于投资未来的不安因素对现在的界定因人而异扔不掉是自己把感情移到物品上面，并因此充分了自己扔不掉=不想扔的机制破烂分三类：1、不用的东西——会念咒的束缚人的淤泥2、还在用的东西——一滩混乱的淤泥3、充满回忆的东西——散发出强大的气场堆积的破烂代表着良心不安的聚集。 9、这些在居住环境里放了好几个月，甚至放了好几年的东西，只不过因为不是生鲜食品所以才没有烂掉。但是如果从机能上来说，它们早已经腐烂了。置身于这样的环境中，就几乎等于是暂住在一个垃圾暂放室。主语到底是“我”还是“物品”？请养成经常如此自问的习惯吧。 10、断舍离是生活的减法哲学，减去多余的物品，认清自我，磨砺感知的本能；俯瞰力是心灵的加法哲学，了解深层需求，相信自我，坚定人生的信念；自在力是人生的乘法哲学，重获生命的自立、自由与自在，解放自我，迈向人生的高境界。 11、就拿断舍离来说，首先出现的是断与舍这种自我肯定、恢
《朗读者》第一季经典语句 第一期·【遇见】 朗读者就是朗读的人，在我看来可以分为两部分来理解，朗读是传播文字，而人则是展现生命，将值得尊重的生命和值得关注的文字完美结合就是我们的《朗读者》。 你有多久没有朗读了很久了吧。因为很多人都觉得朗读，那是学生时代的事情，或者说它只属于一小部分人。朗读属于每一个人。古往今来，有太多太多的文字，在描写着各种各样的遇见。“蒹葭苍苍，白露为霜，所谓伊人，在水一方。”这是撩动心弦的遇见；“这位妹妹，我曾经见过。”这是宝玉和黛玉之间，初见面时欢喜的遇见；“幸会，今晚你好吗”这是《罗马假日》里，安妮公主糊里糊涂的遇见；“遇到你之前，我没有想过结婚，遇到你之后，我结婚没有想过和别的人。”这是钱钟书和杨绛之间，决定一生的遇见。 所以说，遇见仿佛是一种神奇的安排，它是一切的开始。也希望从今天开始，《朗读者》和大家的遇见，能够让我们彼此之间，感受到更多的美好。 1.只要你是个正直的孩子，不管你从事什么行业，你都是我的好孩子。——柳传志 2.如果美丽只是简单地用来做展示，也许它的意义不是那么大，但如果美丽能够转化成一种能力，去帮助更多的人，甚至去让自己变得更好，那它就是很有价值很有意义的了。——董卿 3.愿你慢慢长大，愿你有好运，如果没有，希望你在不幸中学会慈悲；愿你被很多人爱，如果没有，希望你在寂寞中学会宽容。——刘瑜《愿你慢慢长大》 4.生命哪，并不是你活了多少日子，而是你记住了多少日子，要使你过的每一天，都值得回忆。——许渊冲 第二期·【陪伴】 为什么是陪伴其实我们每一期的主题词的选择，是非常慎重的。有时候为了一个主题词会反反复复讨论很长时间，但是陪伴是最早确定下来的主题词，而且从来没有改变过。 我想因为陪伴很温暖，它意味着在这个世界上有人愿意把最美好的东西给你，那就是时间。 当然陪伴也是一个很平常的词，日复一日，年复一年，到最后陪伴就成为了一种习惯，就像我们的朗读者，郑渊洁、乔榛都谈到了自己，夫妻之间的陪伴，父母对孩子的陪伴。 “草，在结它的种子，风，在摇它的叶子，我们俩站着不说话。”在顾城的诗里，陪伴就是这样简单而美好。而在我们每一个人的生命里，会遇到各种各样的陪伴，比如说这会儿，你我之间是一段短暂的陪伴；比如说我们的学生时代，和我们的同学，那是几年的陪伴；还有一种陪伴，是生命里血脉注定一生的陪伴，那是我们和父母，和孩子之间的陪伴。
TEFL New Concept English 1 Well, today ’s story is about a handba啊g.，今天我们来讲一★Pardon?
my 不能单独使用，后面一定要接名词。 mine 后一定不能再接名词。 She\he\it is Lesson 2 Is this your ??
★house There are four people in my family. 在我家里有四口人。 home 抽象的家的概念 说有个人啊，初学英语，半 sorry 呢？他到底做错了什么事呢？
地点副词：（在英语中，时间副词、地点副词的前面不能加介词） one，two ，three，four，five，sir，seven，eight，nine，ten What can I do for you, sir? 某些祈使动词可以后跟and 和另一个祈使动词，而不是后跟带to 的动词不定式结构。Come and see this goldfish. Wait and see. 等着瞧吧。 My ticket is here. 我的票在这。
Lesson 4 Is this your?? 1、代词：he/she/it 的用法。 2、句型：This is.... How many students are there? 4、提出问题：Is Chang-woo Chinese? 看一遍视频，解答问题。 5、精讲课文，板书和笔记。 ’m not sure. So, let Ah, there are six. pictures. 让我们打开书，来数数到底有几个学生。 ★German 德国的，德国人 Germany 德国（国家） He is very nice to his neighbours.
紫雨老师 经典语录 这里的学习紫雨老师经典语录摘录自[自然法则狂热者大厅]紫雨老 师与自然法则学员们交流时的精彩语录 整理：四叶草 版权：紫雨老师 版本号：1.0 build 20080530
前 言 四叶草 日期 ：2008‐5‐30
我学习自然法则有好几个月了。在整个学习自然法则的过程 中，我对自己的人生观点有所改变。而且，心灵经历了一次很大 的解构重组。回想起当时的自己对人生这课题是多么的左右为 难、漂浮不定和充满无力感。这段左右为难的日子困了自己二十 多年，以至于自己失去了对未来的方向。 一向以来，我是父母眼中的好女儿。我从小到大无论做什么事 情都顺从他们的指示。他们说一我不敢说二。因为我以往的经验 告诉我：违逆他们的意思的话，一定会受到非常严厉的责罚。这 也是我自己非常抗拒的，我时常都告诉自己说：“听话照做就好 ，不然就会被骂了！”从小学要参加什么课外活动、中学四年级 要选修什么科、中五毕业之后要继续深造还是出来工作，都已经 被父母内定了。 还记得中五毕业时，我想继续升上中六。但最后我还是服从了 父母的安排，到私立学院去就读电脑高级文凭的课程。用了大概 两年半，就出来社会工作。令我最困扰的时刻其实就是找工作和 应征的时刻。当时的我总是有一种很强烈的“不够” 的感觉——经验不够、学历不够；这个不够，那个不够。。。诸多限制！这 种匮乏的感觉一直维持了三个月，最后才找到了现在这份工作。 刚开始时，我对这份工作感觉还蛮不错。但是随着心境的改变 ，我渐渐对这份工作感到压力。会有压力是因为开始意识到
自己 所喜欢的工作类型，与自己的天赋秉性并不相符。同事们对公司 的批评和责备声浪越来越大，导致我也跟着对这份工作产生了很 强烈的抗拒感。
柳传志经典语录大全 1、企业文化：小企业发展看老板，大企业进步看文化。 2、员工不能把企业当家，家是不努力、有病有灾也能依靠的地方。竞争不讲情面。 3、有5%的希望，就要付出100%的努力。 4、领导不是组织赋予，而是自身魅力展现。 5、看油画的时候，退到更远的距离，才能看明白。离得很近，黑和白是什么意思都分不清。退得远点，就能明白黑是为了衬托白，再远点，才能知道整幅画的意思。打这个比喻是为了时时提醒我们牢记目标，不至于做着做着就做糊涂了。 6、立意高，才能制定出战略，才可能一步步地按照你的立意去做。立意低，只能蒙着做，做到什么样子是什么样子，做公司等于撞大运。 7、我如实跟大家说，我做人画了三个圈，第一个把家里的人，跟我最要好的朋友，这一圈人护好；第二圈，联想的员工，联想的这一圈人，我有这个责任；再大一个圈才是更大的社会。当时间发生矛盾的时候，我非常明白的知道时间怎么摆，精力怎么摆，我实话实说。 8、营销攻略：有5%的希望，就要付出100%的努力。 9、柳传志建议创业者要想清楚三件事情：1.所做的事能不能赚到钱，能赚多少钱、什么时候赚钱；2.行业门槛有多高，进入门槛是什么，自身的水平能否达到行业标准；3.考虑清楚竞争对手的状况，因为竞争关乎生存。 10、最难的地方在于冒风险，冒商业的风险和政策风险。 11、一个公司里面如果没有外边的空降兵进来的话，那就会变得好像近亲繁殖一样。确实有很多新的东西你不了解，也不一定能打开局面，因此有空降兵是必须的。 12、第一代创业的企业家都把工作当事业来做，而一般领导者只是把自己当作职业经理人，或把工作当作谋生的手段，这就有可能出问题，也容易出问题。在基业常青的企业中，怎样让职业经理人把企业当作事业来做，是一个核心问题。 13、我们在选拔人才时要求“德才兼备”，而且“德”一定是放在第一。 14、往上走是代价、是风险，也可能是机会。
Lesson 1 1excuse[ik'skju:z] v.原谅 2me[mi:, mi] pron.我(宾格) 3yes[jes] ad.是的 4is[iz, s, z, ?z] v.be动词现在时第三人称单数5this[eis] pron.这 6your[j?:, j?:, j?r, j?:r] 你的，你们的 7handbag['h?ndb?g] n.(女用)手提包 8pardon['pɑ:d?n] int.原谅，请再说一遍 9it[it] pron.它 10thank you感谢你(们) 11 very much非常地 Lesson 2 1 pen [pen] n.钢笔 2 pencil ['pensl] n.铅笔 3 book [buk] n.书 4 watch [w?t?] n.手表 5 coat [k?ut] n.上衣，外衣 6 dress [dres] n.连衣裙 7 skirt [sk?:t] n.裙子 8 shirt [??:t] n.衬衣 9 car [kɑ:] n.小汽车 10 house [haus] n.房子 Lesson 3 1 umbrella [?m'brel?] n.伞 2 please [pli:z] int.请 3 here [hi?] ad.这里 4 my [mai]我的 5 ticket ['tikit] n.票 6 number ['n?mb?] n.号码 7 five [faiv] num.五 8 sorry ['s?ri] a.对不起的 9 sir [s?:] n.先生 10 cloakroom ['kl?ukru:m] n.衣帽存放处Lesson 4 1 suit [su:t, sju:t] n.一套衣服 2 school [sku:l] n.学校 3 teacher ['ti:t??] n.老师 4 son [s?n] n.儿子 5 daughter ['d?:t?] n.女儿 Lesson 5 1 Mr. ['mist?]先生 2 good [gud] a.好 3 morning ['m?:ni?] n.早晨 4 Miss [mis]小姐 5 new [nju:] a.新的 6 student ['stju:d?nt] n.学生 7 French [frent?] a.& n.法国人 8 German ['d??:m?n] a.& n.德国人 9 nice ['nais] a.美好的 10 meet [mi:t] v.遇见 11 Japanese [?d??p?'ni:z] a.& n.日本人 12 Korean [k?'ri?n] a.& n.韩国人 13 Chinese [?t?ai'ni:z] a.& n.中国人 14 too [tu:] ad.也 Lesson 6 1 make [meik] n.(产品的)牌号 2 Swedish ['swi:di?] a.瑞典的 3 English ['i?gli?] a.英国的 4 American [?'merik?n] a.美国的 5 Italian [i't?li?n] a.意大利的 6 Volvo ['v?lv??] n.沃尔沃 7 Peugeot n.标致 8 Mercedes ['m?:sidi:z] n.梅赛德斯 9 Toyota ['t??j??t?] n.丰田 10 Daewoo n.大宇 11 Mini ['mini] n.迷你 12 Ford [f?:d] n.福特 13 Fiat ['fai?t, -?t] n.菲亚特 Lesson 7 1 I [ai] pron.我 2 am [m, ?m, ?m] v.be 动词现在时第一人称单数
断舍离经典语录100条 断舍离经典语录100条第1段： 1.如果能真的留下必要的物品，那么分类收纳物品之类的技巧也就没什么大的用处了。 2.如果不改变平日的生活习惯，就得不到根治。 3.在研究物品是否应当被留下时，思考的主语是”我”，而不是物品。 4.断舍离，就是透过整理物品了解自我，整理内心的混沌，让人生舒适的行动技术。换句话说，就是利用收拾家里的杂物来整理内心的废物，让人生转而开心的方法。 5.“可惜”不是不用扔东西的赦免令，并且对物品的爱惜之情，要让它成为物品循环的原动力。 6.如果你对别人的东西比对自我的还在意，就说明你对自我太放松，对他人太严格。 7.对人类来说，比起生存需要来，归属认同等需要更为强烈。 8.基本来说，让房间乱七八糟，把屋子搞得一团乱，粗暴地随便乱堆东西，就相当于赋予了自我否定自卑的能量。 9.在断舍离当中，比起凌乱，堆积了聚焦于过去和未来的物品才是问题。 10.不管东西有多贵，有多稀有，能够按照自我是否需要来确定的人才够强大。 11.觉得多长时间属于”此刻”都能够，是每个人的自由；可是必须要有对自我而言最适宜的”此刻”。要是能在
实施断舍离期间，发现最适宜自我的”此刻”，那自然是最好可是了。 12.根本没有必要消除种异样感，如果能够意识到让自我感到异样的原因，所以而卸下包袱，或是对此有所领悟，那么就不会徒增剩余的压力了。 13.断，断绝不需要的东西，舍，舍去剩余的废物，离，脱离对物品的执着，此刻对自我来说不需要的就尽管放手。 14.把物品一次元收拾干净了，以往一向阻碍视听的东西没有了，就能够看得更清楚，就能开启通往更深次元的通道。 15.断，断绝不需要的东西。舍，舍去剩余的废物。离，脱离对物品的执着。此刻对自我来说不需要的就尽管放手。 16.在断舍离里，”扫除”明确地分为上头所说的那种收拾，需要利用收纳术的整理，以及表现为扫擦刷的打扫。三个词的意思几乎是完全不一样的。 17.从加法生活转向减法生活很重要，并不是心灵改变了行动，而是行动带来了心灵的变化，能够说，断舍离就是一种动禅。 18.经过不断地筛选物品的训练，当下的自我就会越来越鲜明地呈此刻自我的眼前，人也就能以此确定出准确的自我形象。 19.断舍离会遏制人类种不知不觉只靠本能与物品打交道的行为。 20.选择物品的窍门，不是“能不能用”，并且“我要不
《朗读者》经典语录 第一期·【遇见】节目卷首语 朗读是传播文字，而人则是展现生命。 朗读属于每一个人。 古往今来，有太多太多的文字，在描写着各种各样的遇见。 “蒹葭苍苍，白露为霜，所谓伊人，在水一方。”这是撩动心弦的遇见； “这位妹妹，我曾经见过。”这是宝玉和黛玉之间，初见面时欢喜的遇见； “幸会，今晚你好吗？”这是《罗马假日》里，安妮公主糊里糊涂的遇见； “遇到你之前，我没有想过结婚，遇到你之后，我结婚没有想过和别的人。”这是钱钟书和绛之间，决定一生的遇见。 所以说，遇见仿佛是一种神奇的安排，它是一切的开始。 只要你是个正直的孩子，不管你从事什么行业，你都是我的好孩子。——柳传志 如果美丽只是简单地用来做展示，也许它的意义不是那么大，但如果美丽能够转化成一种能力，去帮助更多的人，甚至去让自己变得更好，那它就是很有价值很有意义的了。——董卿 愿你慢慢长大，愿你有好运，如果没有，希望你在不幸中学会慈悲；愿你被很多人爱，如果没有，希望你在寂寞中学会宽容。——瑜《愿你慢慢长大》 生命哪，并不是你活了多少日子，而是你记住了多少日子，要使你过的每一天，都值得回忆。——许渊冲 第二期·【陪伴】 我想因为陪伴很温暖，它意味着在这个世界上有人愿意把最美好的东西给你，那就是时间。 当然陪伴也是一个很平常的词，日复一日，年复一年，到最后陪伴就成为了一种习惯，就像我们的朗读者，渊洁、乔榛都谈到了自己，夫妻之间的陪伴，父母对孩子的陪伴。 “草，在结它的种子，风，在摇它的叶子，我们俩站着不说话。”在顾城的诗里，陪伴就是这样简单而美好。而在我们每一个人的生命里，会遇到各种各样的陪伴，比如说这会儿，你我之间是一段短暂的陪伴；比如说我们的学生时代，和我们的同学，那是几年的陪伴；还有一种陪伴，是生命里血脉注定一生的陪伴，那是我们和父母，和孩子之间的陪伴。 “妈妈，你到底为什么爱我？”母亲放下针线，用她的面额，抵住我的前额，温柔地，不迟疑地说：“不为什么，只因你是我的女儿！”——冰心散文《不为什么》 这个世界上绝了哪种生命形式都会导致地球毁灭。狮子和蚂蚁一样伟大，小草和人类一样重要。——渊洁 我要开花，是为了完成作为一株花的庄严生命，不管你们怎么看我，我都要开花！——林清玄《百合花开》 要是没有兔子和鹧鸪，一个田野还成什么田野呢？他们是最简单的土生土长的动物。——梭罗《瓦尔登湖》 第三期·【选择】 生存还是毁灭，这是一个永恒的选择题。以至于到最后，我们成为什么样的人，可能不在于我们的能力，而在于我们的选择。 选择无处不在。面朝大海，春暖花开，是海子的选择；人不是生来被打败的，是海明威的选择；人固有一死，或重于泰山，或轻于鸿毛，是司马迁的选择。 徐静蕾选择了挑战和变化；耶鲁村官玥飞选择了希望的田野；红丝带校长郭小平，选择了呵护与守望；
新概念英语一册L1---L48单词测试 原谅这手提包请再说一遍感谢你非常地钢笔铅笔 书手表外衣连衣裙 衬衣裙子小汽车房子 伞请这里票 号码五对不起先生 衣帽存放处一套衣服学校老师 儿子女儿先生好 早晨小姐新的学生 法国人德国美好的遇见 日本人韩国人中国人也 牌号瑞典的英国的美国的意大利的名字什么国籍 工作电脑键盘操作人员工程师警察女警察出租司机空姐 邮递员护士机械师理发师家庭妇女送牛奶的人怎样今天 身体好美好的谢谢再见 见胖的女人瘦的 高的矮的脏的干净的热的冷的老的年轻的忙的懒的谁的蓝色的大概白色的抓住父亲 母亲女衬衫姐，妹领带兄，弟他的她的颜色 绿色来楼上时髦的帽子相同的可爱的箱子 地毯海关官员女孩 丹麦人朋友挪威人护照 1
棕色的旅游者俄罗斯人荷兰人红色的灰色的黄色的黑色的橘黄色的雇员勤奋的推销员男人办公室助手事情孩子们累渴坐下冰淇淋大的小的开着的关着的轻的重的长的鞋子祖父祖母给 哪一个空的满的尖的钝的瓶子罐头刀子叉子勺子架子课桌桌子盘子食厨香烟电视机地板梳妆台杂志床报纸立体声音响厨房电冰箱右边带电的左边炉子中间房间在哪里客厅靠近窗户扶手椅门图画墙长裤卧室乱使…通风衣服大衣柜掸掉灰尘扫倒空读削尖穿上脱掉开（电灯）关（电灯）花园树 爬谁草横过打字信篮子骨头清洗牙齿（pl.）做（饭）牛奶一顿饭喝水龙头日子云天空照耀家庭步行在…之上桥轮船飞机睡觉刮脸哭 洗等跳照片 2
村庄山谷在…之间另一个 妻子沿着河岸水 建筑物在…旁努力地书架 锤子上漆粉红色最喜欢的作业听碟子前面 仔细的花瓶掉下花 送给带给乳酪面包 肥皂巧克力糖咖啡 茶烟草当然水壶 茶壶沸腾老板分钟 书写糟糕的举起饼干 新鲜的黄油纯净的蜂蜜 成熟的果酱甜的上等的 酒啤酒黑板 L49—96单词测试 卖肉的肉牛肉羔羊肉丈夫牛排肉馅鸡 告诉实情也（否定句）西红柿土豆卷心菜莴苣豌豆豆角梨葡萄桃 希腊气候国家宜人的天气春季有风的温暖的下雨有时夏天秋天冬天下雪一月二月三月4月5月6月 7月8月9月10月 11月12月美国巴西 3
kawayi dog （日）可爱的 you can get in there. I have seen many movies portraying Colombians portray [p?:'trei, p?u-] vt. 描绘；扮演没有欧的音。。O的音 so my question is, why do MANY Americans, like to stereotype that much Colombians, they are warm people stereotype ['steri? vt. 使用铅版；套用老套，使一成不变n. 铅版；陈腔滥调，老套 cocaine [k?药]可卡因 Latino [l?'ti:n?u] n.拉丁美洲人[亦作 latino] 威瑞森（全成威瑞森无线通讯，是美国第一家提供320万像素照相手机配套销售 SIM abbr.subscriber identity module 【电信学】用户身份识别卡 Did I say too much? 我说太多了吗。。。 caption ['k?p??字幕；标题；说明；[英]逮vt. 加上说明；加上标题 1. (新闻报道、文章等的)文首标题；章节标题 2. (插图、图片等的)解说词，说明文字 3. 【电影】字幕
All the foreign films were captioned. 所有外国影片都配了字幕。 I can’t believe you even say that. 牧场调料（Ranch Dressing) 牧场调料的主要成分有蛋黄酱、酸奶和大蒜，其他成分有盐、欧芹、洋葱和辣椒等，其中的酸奶和大蒜赋予撬调料一种特有的冲味儿。白乎乎的牧场调料适合于调拌各种绿色色拉，用它来调拌海味色拉的效果也很好。 Lite Ranch lite低盐的；清淡的；简化的n. 不重要的事物；简化物；软饮料 elite精英；精华；中坚分子韦伯也是按这发音两种都读。。EI是第一种 mayonnaise [,mei?蛋黄酱 lite mayonnaise mustard ['m?st?芥菜；芥末；深黄色；强烈的兴趣 honey mustard蜂蜜芥末
断舍离经典语录 导读：1、物品是一面映照你自己的镜子，它所照出来的是那个你想当做不存在、不愿承认的自己。 2、在断舍离的最初阶段，如果抑制不住地出现"好可惜啊"、"良心不安"之类的感觉，想一想这些"因为不扔东西而造成的损失"，你的感觉就会好很多。 3、断舍离的主角并不是物品，而是自己，而时间轴永远都是现在。选择物品的窍门，不是“能不能用”，而是“我要不要用”，这一点必须铭刻在心。 4、和自己喜欢的东西生活在一起。这样的话就是达到了“断”。 5、收拾，是一种筛选必要物品的工作。在筛选必要物品时候，我们要考虑两个维度，一是我与物品的关系这条关系轴，另一个是当下这条时间轴。换句话说，收拾就是要扪心自问某件物品与当下的自己是不是确实有关系，进而对物品进行取舍、选择的过程。 6、断舍离认为，在扔东西的时候，要把"对不起，谢谢"这样的情绪表达出来。跟那些被你扔掉的东西说说话，做个告别，可以让人更快地整理心情。 7、在避免囤积物品的同时，人对物质的欲望也就淡薄了，反过来，精神世界异常丰富起来。 8、如果能够认为一切物品都是向地球借来的，就能自然而然地涌出感谢与敬畏之情。
9、如果你对别人的东西比对自己的还在意，就说明你对自己太放松，对他人太严格。 10、每天都在用的东西非常容易作用于潜意识。 11、物品其实是物与感情的综合体。即便是同一件东西，自己在这件东西上赋予的感情至关重要。可包含负面感情的物品还是太过沉重了，毕竟我们没必要给我们的人生背上如此沉重的包袱。 12、再接下来就是一些装饰性的给别人看的空间了，这种收纳空间只能放一成东西。减少物品的数量，这样不管是多狭小的旧房子，都能自然而然地营造出高品味的感觉。 13、因此，从今以后也要扔掉那些多余的信息，只选择自己能够付诸行动的信息。尽早从头脑的"便秘"中解脱出来。 14、要成为活在当下，能够立刻付诸行动的人。我想，成功者就是那些能够真正实践的人。 15、我们会在不知不觉中掉进折扣的陷阱，完全忘记了"东西是不是合适自己的品味"。 16、我觉得他们多半是刻意让自己变得没感觉的，虽说不能一概而论，但生活在这种环境里的人，多数都曾经历过强烈的孤独感。寂寞、悲伤的感觉会让他们更痛苦，所以他们干脆封闭了感觉的闸门，让自己麻木不仁。 17、先诊断出物品从你身上掠走了多少能量，然后通过筛选物品的行动，实现自我完善，这就是断舍离的精髓。