MXPA01005631A - Communication device and method - Google Patents

Abstract

A device and a method are proposed in which the sender divides an amount of data to be sent into one or more data units and the receiver acknowledges the correct receipt of data units by returning acknowledgment data units to said sender. The data units are sent by said sender in accordance with a flow control procedure that involves one or more adaptive parameters. After a given data unit is sent, the sender performs a data loss detection routine, and ifa triggering event occurs, a corresponding response procedure is conducted, where this response procedure comprises at least two different modes for adapting said one or more adaptive parameters. Preferably, there is a first and a second mode, where the first mode is associated with the actual loss of a data unit, so that this first mode comprises the conventional data loss procedures, and where the second mode is associated with the recognition that an excessive delay has taken place and not a loss of a unit.

Description

DEVICE AND COMMUNICATION METHOD The present invention relates to a communication device and method wherein a communication oriented to the unit of data between one sending and a receiver is carried out by the sending and the receiver operating in accordance with a predetermined protocol. The communication oriented to the data unit it is well known In the data unit oriented communication a quantity of data is divided into one or more data units, where the structure of the data units is defined by a communication protocol to which the sending and receiving adhere. the receiver in the communication The protocol also defines how specific information is to be encoded, and how the sender and / or receiver can react to specific information. Unit-oriented communication is also known as packet exchange communication. It should be noted that the data units used in relation to specific protocols have different names, such as packages. frames, segments, etc. Fara the purpose of the present description, the term "data unit" generically ST will refer to all types of units used in a - communication oriented to data unit. A particularity that many communication protocols use to increase reliability is that of recognizing the data received. More specifically, one sending or transmitting port of the given protocol sends data units, and the receiving or receiving port of the given protocol recognizes the correct receipt to return appropriate knowledge data units. In this way, the transmission port is informed that the data units that were sent were also received correctly, and consequently can adjust consequently the flow control of the additional data units to be sent. An example of a protocol that uses knowledge data units is the so-called transmission control protocol (TCP), which is a part of the TCP / IP protocol series The transmission control protocol and the TCP / IP protocol series they are well described, eg, in '' TCP / IP Illustrated, Volume 1 - The Protocola- "by W. Richard Stevens, Addison-Wesley, 1994. In order to come across the fact that the data units or knowledge data units may be lost, a particularity of time out is provided in many protocols, said particularity d? Time out means that a period of time out is adjusted when the data is sent, and s. the specific data has not been recognized by time, the time period outside expires, a time out response procedure is initiated In TCP. the time-out response consists of retransmitting the data that was not recognized, and readjusting one or more flow control parameters. As an example, TCP uses a window-based flow control. TCP is a byte-oriented protocol that divides a certain number of bytes to be sent in so-called segments and a record of the data sent is kept in terms of bytes , that is, until which byte the data was sent and a record of the received data is also kept in terms of bytes, that is, until the data was received byte. The simplest way to control the flow of segments in connection with messages with knowledge would be to send a segment and not send the next segment until the last segment sent was recognized. Such method of flow control, however, would not be very efficient As already mentioned, TCP uses window-based flow control, which is also referred to as flow control in accordance with sliding windows. This concept is also well described in the book mentioned above by w, Richard Stevens Figure 2 illustrates the concept of sliding windows. As you can see, an amount of 8,192 bytes is going to be sent in the example, where this quantity is divided into 8 segments. Did you send segments? it controls in accordance with the sending window, where the left end of the sending window is defined by the data in the segments that have been sent and have already been recognized. In the example of Figure 2, this TS? L data up to 2,048 bytes, that is, segments 1 and 2. The adjustment of the length of the sending window, and in this way the right end of the window is a matter of the control procedure, which does not need explain in detail in this - * - * The send window defines the amount of data that may have your corresponding knowledge pending. In the example in Figure 2, the data up 4 096 bytes, that is, segments 3 and 4 have been sent and not yet recognized, and the difference between said 0 sending and the unrecognized segments and the right end of the window for sending the fine usable window, that is, the Data may still have to be sent without having received any additional knowledge. As a consequence, in the example of Figure 2, segments 5 5 and 6 can still be sent, but segments 5 and 8 they can only be sent if the window moves to the right, what happens if the additional segments are recognized so that the left end moves to the right and / or if the length of the transmission window increases, In addition, it should be noted that TCP provides cumulative recognition, that is, there is no one-to-one correspondence between segments and recognitions by segments, because a recognition message can cover a plurality of segments. As an example, the port of reception for the amount d? The data shown in Figure 2 could send a byte recognition up to 4.096, so that this recognition message would cover both segments 3 and 4. The transmission window used by the transmission port will typically be determined by the so-called offered or advertised window, which is a length of data provided to the port of delivery by the port of reception. In this way. the reception port can influence how many segments the transmission port will send at a time, and typically the advertised window will be calculated on the basis of the reception buffer of the reception port, Also, the advertised window is a dynamic parameter that can be change with every acknowledgment sent by The reception port Beyond the public window, it is also known how to define the congestion window call, which ST uses in connection with several congestion control routines such as slow start, avoid congestion. rapid retransmission and quick recovery, again see for example the book mentioned above by W. Richard Stevens. The congestion window is a record that holds the shipping port, and is intended to take into account the congestion along the connection? Ntr © the shipping port and the receiving port. As a typical control mechanism, the send window will be defined as the lowest of the advertised window and the congestion window. Even when the advertised window is a control d? flow imposed by the reception port, the window d? Congestion is a flow control impulse by the shipping port. as a mechanism to take into account congestion. In a general sense, the congestion window is an example of an adaptive flow control parameter. In TCP, the above-mentioned out-of-time response consists in readjusting the congestion window to a segment and then consequently sending only one component, to say, retransmit the segment that was not recognized and thus caused the time out, The sending port then waits for the recognition of the retransmitted segment Another example of an adaptive flow control parameter is the time period itself, which, v.gr . in TCP it is referred to as RTO (Time Out of Retransmission), The RTO is duplicated as a response to time out. As already mentioned, the particularity of time is a mechanism for detecting data loss. Other mechanism of data loss detection-exist. Another example is the retransmission of data units in TCP in response to the receipt of duplicate acknowledgments. This mechanism will be explained briefly in what follows. As already mentioned (see, v.gr .. Figure 2). A quantity of data to be sent is divided into a sequence. Conventional implementations of TCP are arranged so that if the receiving port has received and recognized a certain amount of data up to a given byte (a certain number of consecutive segments), it expects data to be next in the sequence. For example, if segments have been received up to segment 4, segment 4 is recognized and the receiving port expects to receive segment 5. If then it receives an additional data unit that is different from the segment 5 (v gr, segments 6, 7 and 8), continues to the recognition segment 4 for each data unit it receives. As a consequence, the transmission port receives duplicate acknowledgments. Commonly, TCP is implemented in such a way that the port d? transmission receives duplicate acknowledgments Commonly, TCP is implemented in such a way that the transmission port will count the number of duplicate acknowledgments and if a certain threshold value is reached (eg, 3), then the next unit of data in the sequence to the data unit for which the duplicate acknowledgments were received, it is retransmitted. The object of the present invention is to improve the communication in a system using a communication protocol that specifies the recognition of sent data and specifies a data loss detection function, such as a time out function or a response function of duplicate recognition. This object is solved by a method as described in claim 1 and a device as described in claim 18. In accordance with the present invention, one who sends in a communication will conduct a procedure d? response in response to an event that triggers a data loss detection mechanism, wherein the response procedure comprises at least two different modes for adapting the adaptation parameters used in the flow control. In this way, the method and device of the present invention are highly flexible in their handling of percussion events and can be implemented especially in such a way that the response procedure can be selected depending on several potential causes in the percussion event, so that the correct response measures can be invoked to a given situation, and from this Ways you can avoid measures that could actually aggravate situations that may occur after you have triggered a data loss detection mechanism The TS data loss detection mechanism is a mechanism that is able to detect a loss of data. The examples are a time-out mechanism or a duplicate recognition mechanism. Naturally, the invention can be applied to any appropriate data loss detection mechanism. In accordance with the invention, a response procedure comprises at least two different modes to adapt the adaptation parameters used with flow control. As a example, which constitutes a preferred modalidac there are two modes, which are respectively associated with different causes of a time out or a predetermined number of duplicate acknowledgments (v gr the 3 mentioned above) More specifically, a first mode is associated with the loss of a data unit and the second mode is associated with an excessive delay along the connection. Due to the use of two different modes, TS can adapt the parameters as appropriate for the cause of time out or duplicate acknowledgments-consequently, the flow control procedure will contain one or more evaluation and judgment steps in which the event is qualified. percussion gr, a category formation is conducted as to what caused the event Then, depending on the result of this characterization, an appropriate response procedure can be trained In the context of the previous example, if it is determined that the time out or the Duplicate acknowledgments are caused by the loss of a unit of data, then the known response procedure to the loss of data units can be run vg, as is known from conventional TCP, which assumes that any time out or the recurrence of several Duplicate acknowledgments are caused by the loss of the data unit In accordance with the In this mode, however, there is a second mode and if it is determined that the time out or duplicate acknowledgments are caused by an excessive delay along the connection, then you run a procedure d? excessive delay response, which will typically be different from the response procedure to the loss of a data unit More specifically, as will also be explained in more detail in the following, the judgment that the data units have been lost will be answered by reducing the transmission scheme to thereby avoid additional congestion On the other hand, if there is an excessive delay along the connection, then the measures taken in response to a supposed loss of data units would not be helpful, rather they could actually aggravating the problem causing the excessive delay Consequently the response procedure the excessive delay will typically be different and e.g., comprises maintaining the transmission rate at the previous level, but on the other hand increasing the time period outside, so that additional unnecessary retransmissions are avoided. Naturally, the present invention can be implemented propo by providing an arbitrary number of modes or procedures for responding to various causes of percussion events. The number of modes and the specific measurements taken in each mode naturally depend on the specific situation. that is, the selected protocol, the communication situation determined, etc. An important aspect of the present invention is that even when the data loss detection mechanism is capable of detecting data loss, the reaction to the percussion of the data loss detection mechanism does not assume that data loss has occurred. necessarily, a flexible response is possible, which can take into account various causes of the percussion event. Additional aspects and advantages of the present invention will be better understood from the following detailed description, which refers to the figures, in which: Figure 1 shows a preferred embodiment of the control method according to the present invention; Figure 2 is an explanation diagram to describe the concept of window-based flow control; Figure 3 is a graph to explain the advantages of the present invention; Y Figure 4 is an explanatory diagram to illustrate a situation in which an excessive delay can be caused in a connection between two host computers. Although the following description will generally address any communications protocol that makes use of data recognition and also provides a particularity of time out frequently examples will be provided that relate to the known TCP transmission control protocol of the TCP / IP protocol series The application of the present invention to this protocol is a preferred embodiment In order to avoid any unnecessary repetition the Exposure in the introduction to this application is incorporated into the invention exhibition. Figure 1 shows a partial flow diagram of a preferred embodiment of the present invention. As can be seen, the pauo SI indicates that a response procedure is introduced. Figure 1 does not show the flow control procedure leading to this point, since which is not of importance for the present invention. For example, it may be the flow control procedure based on the explanation window in relation to Figure 2 and well known v.gr of TCP. It is only important for the invention that there is data recognition and a particularity of data loss detection, such as a protocol transmission port has the capacity d? detect a possible or potential loss of data, and can conduct a corresponding response procedure. As already mentioned, the particularity of detection of data loss, eg, may be a particularity of time out or a particularity of detection of duplicate recognition. In the example of Figure 1, after the response procedure is entered, the parameters d? selected adaptation that are used for flow control and are stored and then readjusted to values determined in step S2. As an example, the time period outside and / or the congestion window described above are said adaptation flow control parameters. In conventional TCP, the congestion window is typically reset to a segment value and at the same time RTO is duplicated. It should be noted that not all adaptation parameters used in the flow control procedure in fact need to be changed, much more only a selected number. Also, it should be evident that the present invention is naturally not restricted to the control of window-based flow and the associated adaptation parameters, but rather the invention is applicable to any principle of flow control and the associated adaptation parameters. Returning to Figure 1. the data unit that triggered the event (v gr, caused a time out) is retransmitted in step S3 In other words, when it stays with the example of a time out, the data unit for which no acknowledgment was received during the timeout period is retransmitted. Then, at a later point it is determined in step S4 whether a recognition associated with the retransmitted data unit has been received. This may be an acknowledgment. cumulative and also a simple recognition It may be noted that the dashed lines in Figure 1 indicate that other steps may be interposed, but these are not of importance for the present invention Then, in accordance with the preferred example of Figure 1, step S5 determines whether the recognition associated with the data unit that was retransmitted actually acknowledges the original transmission of the data unit or the retransmission It should be noted that the "original transmission" may already be a retransmission, so that "retransmission" may be the retransmission of a retransmission, etc. various possibilities to implement step S5. as explained later. If step S5 determines that the recognition message does in fact acknowledge the retransmission of the data unit, then the procedure goes to step S7, in which a loss response procedure d is run? data unit, because the negative result of step S5 of decision indicates that the original transmission of the data unit was lost. In the example d? TCP, step S7 will consist of conventional measures against loss of data unit. On the contrary, if step S5 of decision is answered in the affirmative, then the procedure goes to step S6, in which a response procedure is run which answers an excessive delay. In other words, because step S5 indicated that in fact the original transmission of the data unit was not lost, but only excessively delayed, the corresponding measures must be taken. For example, when TCP is taken as an example of a protocol, this may consist of returning the congestion window to the value stored in step S2 and on the other hand adapting the time period to the delay In other words, the time of travel RTT round associated with the original transmission and recognition of the Original transmission can be used as a basis to adapt the time period out. In this way, retransmissions and extra timeouts or unnecessary duplicate recognitions due to excessive delay can be avoided. Preferably, the congestion window is not simply readjusted to the previous value, but rather adjusts to the value it would have assumed, if the response procedure had not occurred, that is, if the data loss detection mechanism had not triggered. . As can be seen, the example of Figure 1 shows a first mode consisting of steps S2, S3, S4, S5 and S7, as well as a second mode consisting of Ios-steps S2. S3, S4, S5 and S6 In order to better explain the present invention, reference will now be made to Figure 3, which shows an example of a flow control procedure conducted in connection with conventional TCP. The graph shows the amount of data in bytes carried over time. As you can see, the first two segments are sent as time t = 4s. Then, due to the interaction of receiving recognition data units and adjusting adaptation parameters not shown, segments are sent.

For the purpose of explanation it should be noted that the symbols of diamond shape refer to segments, and the square symbols to units d? Recognition data The diamond symbols indicate the first byte of the segment, while the squares indicate the lowest unrecognized byte The recognition data units indicated in a certain segment level always recognize the segments sent up to the segment level In other words, the recognition at the 6,400-byte segment level (t = 12s) recognizes the segments below 6,500 byte, but not including the 6,400 byte. Quite the contrary, as explicitly indicated on the graph the segment at 6,400 byte (t = 10s) is a data unit or packet that causes a time out. As a consequence, a retransmission is conducted from the data unit at the 6 400 byte level. Now if it is assumed that the time was shown in Figure 3 it was caused by an excessive delay and not for the first shown packet that is lost, then retransmission has the following negative consequences. For one thing, it leads to a general decreased operation, since the same data has to cross the connection or the double connection path that wastes bandwidths that could otherwise have been used for useful data This negative consequence will occur in any protocol that responds falsely to a time out by retransmitting the data unit. Yes, as shown in Figure 3 the TCP protocol is used, then the reaction of the transmission port to said time out not caused by loss of data unit is particularly disadvantageous the sender will retransmit all the pending packets and besides that will reduce its transmission regime. This is shown explicitly in Figure 3 It can be seen that the time out above described not caused by loss of the data unit is also referred to as a time out for Lto As shown also in Figure 3, in conventional TCP sending it misinterprets all acknowledgments associated with the retransmitted data units as retransmission recognition, even though these acknowledgments (ACKs) are in fact delayed acknowledgments of the original transmissions. What Figure 3 does not show, is that in addition to the duplicated data units sent by the port The transmission will trigger duplicate acknowledgments at the reception port, which will lead to yet another reduction in the transmission rate at which it sends conventional TCP, that is, the congestion window adjusts to half its previous value. The occurrence of excessive delay that goes beyond what the time period outside of TCP may consider, may appear in wireless networks or such protocol connections of which at least a part runs through a wireless link. The inventors of the present application observed that out-of-time times may occur sufficiently frequently in such networks, so that operation degradation would be serious. Examples of this will now be briefly mentioned. Figure 4 shows a situation, where two guest computers act as TCP ports (indicated by the long arrows from guest to guest in the background and upper part of the figure). The lower protocol layers comprise a radio link through a wireless access network to the internet. The connection between the internet and the guest on the right is not shown. An example of a protocol for the radio link is the so-called RLC radio link control protocol. As indicated in Figure 4, both the transport layer protocol (e.g., TCP) as the link layer protocol (eg, RLC) have an AR (Automatic Retransmission Request) function. This means that these protocols will implement both time-out and retransmission functions. In the situation of Figure 4, because the ARQ is being used in the link layer, a race condition is generated between the link layer and the layer. Transport: While the link layer retransmits data, the transport relay timer may expire, leading to a parasitic time out. Retransmissions in the link layer may be due, eg, to transmission errors or loss of data due to passes. It can also be seen that the transmission delay over the wireless network is frequently a considerable fraction of the end-to-end delay between the transmit and receive port of the transport layer protocol. If in this case the available bandwidth to the transport layer connection in the wireless network drops considerably over a short period of time, the resulting increase in the end-to-end delay between the sending and the layer receiver Transportation can lead to out-of-time parasites. The examples of Bandwidth drops include mobile hosts running a pass to a cell that provides less bandwidth than the old cell As previously indicated when the present invention is employed, the problem described in relation to Figure 3 can be avoided More specifically when the descptc method is applied in connection with Figure 1 to the problem in Figure 3, then the transmission port is able to distinguish between the recognition data units to the original transmission of a data unit and data units. recognition of the retransmission of a data unit From this information, the transmitter can decide whether a parasite timeout has occurred or if there has actually been a loss of a data unit. The transmitter can then react accordingly More specifically, in the example of Figure 3, the transmitter using the invention is capable of identifying the unit d of recognition data received after having retransmitted the first packet shown co or being an acknowledgment for the original transmission (t = 10s) and not for retransmission (t = 15sj Because of this, the transmitter will perform an appropriate response procedure for the excessive delay to say not to retransmit the data units after first data unit retransmitted and also not decrease the transmission rate but rather the transmitter will increase the time period used in the flow control on the basis of the delay measured between the original transmission of the data unit and the receipt of the corresponding recognition data unit for the original transmission In this way, retransmissions and out-of-time times can be avoided As can be seen The present invention is capable of providing a mechanism that allows a more flexible communication system when a protocol providing data recognition and a time-out function or function of detecting data is used. duplicate recognition In the example described, the invention is capable of calibrating a percussion event, that is, distinguishing between at least two different causes and then being able to invoke an appropriate response procedure. It can be seen that in the previous examples the modes for adapting the parameters of adaptation were associated with loss of data unit on the one hand and excessive delay on the other but of course the present invention is in no way restricted to this. Rather the moaos for adapting the adaptation parameters can be ascertained with any possible cause of time-out events or duplicate recognition events In the embodiment described in Figure 1. it was decided in step S5 whether the recognition data unit associated with a given data unit recognized the original transmission or the retransmission of the determined data unit In accordance with a first preferred embodiment for implementing this step, the transmitter maintains a record of a round trip time FTT associated with the connection between the transmit and receive port and especially keeps a record of the shortest RTT found during the connection or the session up to the point of time or consideration Then if a recognition data unit for a retransmitted data unit is received within a period of time that TS less than a predetermined fraction of the shorter RTT, then the transmitter determines that this recognition belongs to the normal transmission and not the retransmission This fraction can be adjusted to a specific value or can be an adaptation parameter. It is naturally not necessary for the comparison value multiplied with the fraction to be the shortest RTT measured. It is also possible for the transmitter to maintain an average RTT value. In this sense the value of The comparison to multiply by the fraction is usually a function of one or more RTT values measured in the course of the connection (during the session). In accordance with another preferred embodiment for implementing step S5, the transmitter adds a mark to sending data units, wherein the mark is defined in such a way as to distinguish between an original transmission and a retransmission. Then, the receiver can consistently mark the recognition data units, so that the transmitter is able to identify whether a recognition refers to the original transmission or retransmission. This marking of data units can be done in any desired way. for example, in theory it would be possible to simply designate a single bit in the data units, where a value of 0 would indicate the original transmission and a value of 1 a retransmission, or vice versa. In a general sense, a bit string can be selected so that it can also transfer some more information. Nevertheless. in connection with protocols that provide said option, it is preferred to use the time stamp option. This option e.g., it is well known for TCP, see the book mentioned above by W. R. Stevens. In other words, it is preferred to include a time stamp in the units of sent data, which indicates when the data unit was sent. The receiver can then simply include the same time stamp in the knowledge data unit, so that the transmitter has a unique way of identifying the data units to which the recognition relates. Although the present invention has been described in relation to preferred embodiments, they do not restrict the scope, and are only intended to convey a better understanding of the invention. Rather, the scope of the invention is determined by the appended claims.

Claims (4)

1. CLAIMS 1. - A method for controlling a communication oriented to a data unit between a transmitter and a receiver that operates, according to a predetermined communication protocol, wherein the transmitter sends a quantity of data to be sent to one or more data units that have a structure determined by the protocol, the receiver acknowledges receipt of the correct data units by returning knowledge data units to the transmitter, the data units are sent by the transmitter in accordance with a flow control procedure conducted on the basis of of one or more parameters d? adaptation and knowledge data units, and the flow control procedure comprises a data loss detection mechanism capable of detecting data loss in communication, the data loss detection mechanism being triggered to indicate potential loss of data by one or more predetermined events, wherein in response to the percussion of the data loss detection mechanism a corresponding response procedure is conducted, the response procedure comprising at least two different modes to adapt to one or more adaptation parameters.
2. 2. A method according to claim 1, wherein the TS data loss detection mechanism is a time-out mechanism, so that after a data unit is sent, the transmitter monitors a period of time out. and if no knowledge data unit associated with the data unit is received before the timeout period expires, the timeout mechanism is triggered.
3. 3. A method according to claim 1, wherein the mechanism for detecting and losing data is a duplicated knowledge detection mechanism, d? so that the transmitter monitors the received knowledge, and if a data unit is recognized a predetermined number of times, the duplicate recognition detection mechanism is triggered.
4. 4. A method according to claim 2 or 3, wherein the response method comprises the retransmission of a determined data unit. 5 - A method according to claim 4, wherein the decision about which of the at least two modes is selected to adapt the adaptation parameters is made on the basis of one or more units of recognition data received by the transmitter after having retransmitted the determined data unit 6 - A method of compliance with the claim 2, wherein the time-out period is one of the adaptation parameters. 7 - The method of compliance with one of the 10 claims 1 to 6, wherein the flow control procedure is window based, and one or more flow control windows are among the adaptation parameters 3 - A method according to claim 5, wherein said when less two modes consist of a first mode and a second mode, the first mode being associated with the judgment that the percussion event was caused by the loss of a given unit of data, and the The second mode is associated with the judgment that the given data unit or the knowledge data unit for the given data unit have been excessively delayed. 9 - A method of compliance with p c. claim 8, wherein the transmitter brand data units that are being sent so that an original transmission can be distinguished from a retransmission, and the receiver correspondingly marks the knowledge data units, so that the knowledge of an originally sent data unit can be distinguished from the knowledge of retransmission of the data unit 10 - A method according to claim 8, wherein the transmitter marks the data units including a time stamp in each data unit sent, the time stamp indicating the time in which said unit was sent d? data, and the receiver marks the knowledge data unit for a received data unit including the time stamp contained in the received data unit in the knowledge data unit for the received data unit 11 - A compliance method with claim 9, wherein the transmitter dials the data units including a bit string in each data unit sent, the bit string having at least two different values to distinguish between an original transmission and a retransmission, and the receiver dials The knowledge data unit for a received data unit includingthe bit string contained in the data unit stored in the knowledge data unit for the received data unit 12 - A method according to claim 11, wherein the bit string includes a single bit 13 - A method of according to claim 11, wherein the bit string consists of a plurality of b ts so that the bit string is able to distinguish between different retransmissions 14 - A method according to one of claims 10 to 13 wherein the The first mode is selected if the first unit of knowledge data associated with the determined unit of data that is received after having retransmitted the determined data unit recognizes the retransmission of the determined data unit and the second mode is selected in the first unit. data-knowledge associated with the given data unit that is received after having retransmitted the given data unit recognizes the transmitted n original of the determined data unit 15 - A method of conformity in claim 8, wherein the transmitter measures the round trip time associated with the connection to send the amount of data, the time between the retransmission of the determined data unit and the receipt of the first knowledge data unit associated with the determined data unit is determined and compares with a value derived from one or more of the round trip time measurements, and the first or second mode is selected based on the result of the comparison. 16. The method according to claim 15, wherein the value derived from round trip time measurements is the shortest round trip time for the connection, and the second mode is selected if the time between the retransmission of the determined data unit and the receipt of the first knowledge data unit associated with the determined data unit is less than a predetermined fraction of the travel time. round minor. 17. A method according to one of claims 8 to 16, wherein the second mode comprises adapting the time period out based on the time that elapsed. between the original transmission of the determined data unit and the receipt of the first knowledge data unit associated with the determined data unit 18 - A method according to one of claims 8 to 17, wherein the control method of flujc is window-based and a congestion window is used where the value of the congestion window at the time of the percussion event is stored afterwards. 10 the percussion event occurred and subsequently the value of the congestion window is reset to a predetermined value, and if the second mode is selected after having received the first knowledge data unit associated with the determined data unit 15? value of the congestion window is adjusted to the value would have been assumed, had the response procedure not occurred. 19- A communication device for communication oriented to conformance data unit 20 with a predetermined communication protocol, wherein the communication protocol prescribes that the sender in a communication divides a quantity of data to be sent in a communication. or more data units that have a structure determined by the protocol and the receiver p in the communication recognizes the correct receipt of the data units returning knowledge data units to the transmitter, wherein the communication device, when acting as a transmitter is arranged to send data units in accordance with a flow control procedure that is driven on the basis of one or more adaptation parameters and knowledge data units, the flow control method comprising a detection mechanism for data loss capable of detecting data loss in communication, the data loss detection mechanism being triggered to indicate the potential loss of data by one or more predetermined events, wherein in response to the percussion of the data loss detection mechanism, a corresponding response procedure is conducted, the response method comprising at least two different modes for adapting to one or more adaptation parameters 20. A device according to claim 19, wherein the data loss detection mechanism is a time-out mechanism, so as to that after a unit of data is sent, the device when acting as a transmitter supervises a period of time out and if it does not receive any knowledge data unit associated with the data unit before the time period expires, the time out mechanism is triggered. 21, - A device according to claim 19, wherein the data loss detection mechanism TS a duplicate knowledge detection mechanism, so that the device when acting as a The transmitter monitors the knowledge received and if a data unit is recognized a predetermined number times, a duplicate knowledge detection mechanism is triggered. 22. A device according to one of claims 19 to 21, wherein the response method comprises the retransmission of a determined data unit. 23. A device according to claim 22, wherein the decision on which of the at least two modes to select to adapt the adaptation parameters is made on the basis of one or more units of knowledge data received by the transmission after having retransmitted the determined pc data unit. 24. - A device according to claim 20, wherein the time period out is one of the adaptation parameters. - A device according to one of claims 19 to 24, wherein the flow control method is based on a window, and one or more flow control windows are among the adaptation parameters 26 - A device according to claim 23, wherein the at least two modes consist of a first and a second mode, the first mode being associated with the judgment that the percussion event has been caused by the loss of the determined unit of data and the second mode being associated with the judgment that the determined unit of data or the unit of knowledge data for the unit of determined data has been excessively delayed 27.- A device according to claim 26, wherein the device when acting as a transmitter marks unidaces of data that is sent so that an original transmission can be distinguished from a retransmission and the device when acting as a corresponding mark receiver the knowledge data units of So that the knowledge of a data unit originally sent can be distinguished from the knowledge of the retransmission of the data unit. 28. A device according to claim 27, wherein the device when acting as a transmitter marks the data units including a time stamp in each data unit sent, the time stamp indicating the time in which the unit of data. data was sent, and the device when acting as a receiver marks the knowledge data unit for a received data unit including the time stamp contained in the received data unit in the knowledge data unit for the received data unit . 29. A device according to claim 27, wherein the device when acting as a transmitter marks data units including a bit string in each data unit sent, the bit string having at least two different values to distinguish between an original transmission and a retransmission, and the device when acting as a receiver marks the knowledge data unit for a received data unit including the data string contained in the received data unit in the unit of knowledge data for the unit from received data. 30. A device according to claim 28 or 29, wherein the first mode is selected if the first knowledge data unit associated with the determined data unit that is received after having retransmitted the determined data unit recognizes the retransmission of the determined data unit, and the second mode is selected if the first knowledge data unit associated with the given data unit that is received after the determined data unit has retransmitted recognizes the original transmission of the data unit determined. 31. A device according to claim 26, wherein the device when acting as a transmitter measures the round trip time associated with the connection to send the amount of data, the time between the retransmission of the determined data unit and the receipt of the first associated knowledge data unit. with the determined data unit is determined and compared with a value derived from one or more of the round trip time measurements, and the first and second modes are selected on the basis of the comparison result 32 - A device according to claim 31, wherein the value derived from round trip time measurements is the shortest round trip time for the connection, and the second mode is selected if the time between the retransmission of the determined data unit and the receipt of the first knowledge data unit associated with the given data unit is less than a predetermined fraction of the shortest round trip time 33 - A device in accordance with a of claims 26 to 32 wherein the second mode comprises adapting the time period off on the basis of the time that elapsed between the original transmission of the determined data unit and the receipt of the first knowledge data unit associated with the determined data unit 34 - A device according to one of claims 26 to 33 wherein the flux control procedure or is window-based and a congestion window is used where the value of the congestion window at the time of the percussion event is stored after the percussion event occurred and subsequently the value of the congestion window is reset to a predetermined value, and if the second mode is selected after having received the first knowledge data unit associated with the given data unit, the value of the congestion window is adjusted to the value would have been assumed, if the response procedure had not occurred RESOLUTION OF THE INVENTION A device and a method are proposed in which the sender or transmitter divides a quantity of data to be sent to one or more data units and the receiver recognizes the correct receipt of the data units by returning recognition data units. to the transmitter. The data units are sent by the transmitter in accordance with a flow control procedure that involves one or more adaptation parameters. After a given data unit is sent, the transmitter performs a data loss detection routine, and if a percussion event occurs, a corresponding response procedure is conducted, wherein this response procedure comprises at least two modes different to adapt to one or more of the adaptation parameters. Preferably, there is a first and a second mode, wherein the first mode is associated with the actual loss of a data unit, so that this first mode comprises the conventional data loss procedures, and wherein the second mode is associated with the recognition that an excessive delay has occurred and not a loss of a unit.