System and Method for Multi-Channel Packet Transmission The present invention relates to packet transmission. In a variant, the invention relates to packet radio transmission. BACKGROUND ART Some wireless communication systems are based on packet radio paradigm. An example of such systems is the so-called Ethernet radio systems. Such packet radio systems typically require transmission schemes with the following functionalities: - exploit as fully as possible the available bandwidth of point-to-point multi-link radio equipments; - guarantee as much as possible a high level of quality of service (QoS), traffic aggregation and drop precedence; - dynamic admission control of packets or frames; - fast restoration of traffic in case of hardware failure; - no, or insignificant impact on traffic in case of single or multiple radio failure or multi-link throughput change. SUMMARY In view of the above requirement, a new approach is desired in designing the architecture of new generation point-to-point multi-link wireless systems. The known protection schemes available on legacy radio equipments {e.g., N:1 , N+1 or 1+ 1) or legacy multilink configuration (such as N+0) typically require that packets are statically mapped into a single link. However, in order to achieve as much as possible a complete load balance in a multi-link system and a per flow/conversation protection engine is desirable to preserve some kind of priority hierarchy among incoming services in most or all conditions. It is further desirable for such engine to operate in real-time and in a transparent manner preferably without disrupting served conversation/flow. In radio applications, due to the limited availability of spectrum, it is not desirable to waste radio resources (bandwidth). Therefore, an engineered traffic distribution mechanism is typically required that allows for fully (or at least efficiently) exploiting the available resources. Accordingly an intelligent per flow/conversation protection scheme that is autonomously and dynamically capable of discarding low priority flows/conversations and preserve high priority flows/conversations is desirable. Moreover, with the introduction of Adaptive Coded Modulation (in which the modulation, coding and possibly other parameters are typically adapted according to the conditions present on the radio link), it becomes desirable that traffic aggregation and packet protection be performed with a solution capable of overcoming, or at least substantially obviating, the drawbacks of the static approach which is a solution typically provided by some legacy radio systems. Whenever radio bandwidth changes due to radio propagation condition, it is desired that the traffic impact be limited as much as possible, improving the protection efficiency and potentially approaching a 'hitless' condition. The more limited the availability of transmission resource (bandwidth) becomes, such as in digital radio transmission systems, the more desirable it becomes to make effective usage of the bandwidth available from parallel transmission resources in a radio multi-link transmission system for packet based traffic. In addition to the above-described scenario, another variable which may need to be kept under control is the latency of the system. Using one channel at a time for the traffic transmission, the delay experienced packet by packet will typically be the one given by the single link. On the contrary, by considering the entire multi-link array as a single Virtual Link, latency may be reduced to improve the performance of the system. Some solutions addressing the above problem are already known. One of such known solutions is based on standard Link Aggregation known as LAG, IEEE 802.1 AX-2008. A non-exhaustive, brief description of some of the functionalities proposed by LAG, IEEE 802.1 AX-2008 (for simplification hereinafter referred to LAG) is given below: The load balancing that can be achieved in LAG is typically strongly dependent on the statistics of the traffic to be spanned over the multi-link transmission system. Typical implementations are based on hashing algorithm that distributes the traffic over the available channels in a random way. The hashing function typically operates on the content of some specific/standard fields of the packets themselves. This approach may not guarantee an effective usage of the total bandwidth available and the results may be even poorer if the traffic is encrypted with mechanisms such as IP-SEC (standing for Internet Protocol Security, which relates to providing security to IP traffic using authentication and encryption on IP packets). Standard LAG typically addresses physical layers with high reliability {e.g. fiber or copper). The operating status of such physical layers can typically be easily referred to as 'working' or 'not-working' which may correspond to 100% or 0% availability of bandwidth. A change in such status typically happens quite seldom as it can be mainly due to failure {e.g., hardware failure) of the equipment or the line. As a consequence, when a link changes to 'not-working' status, an impact on the traffic can be accepted only because it is coming from failure which is supposed to be a scarce event. On the contrary, when the physical layer is radio with adaptive modulation, the status may correspond to 100% bandwidth availability or less {e.g. 75%, 50%, 25% or 0%) and the change of such status may happen more often as it may be due to radio propagation conditions, e.g. due to weather conditions (which is a normal and expected behavior of radio media). • When one of the links becomes unavailable, the total throughput is then typically reduced: standard LAG typically does not guarantee only the dropping of a low priority traffic. A typical scenario may be the following: LAG foresees a distributor spanning incoming conversation over a set of physical interfaces. Conversations are distributed regardless of their traffic/service type or priority. Once distributed, QoS mechanism applies only on the physical interface and not on the aggregated traffic. In such a way, if physical interface is overloaded with all high priority traffic, even the high priority traffic may be discarded. At the end, LAG first distributes, then deals with priority having QoS on physical interfaces. Standard LAG typically does not allow spreading conversations over multiple links at the same time, because reordering is not contemplated. This leads to a not guaranteed load balance. When conversations have different bandwidth profiles, allocating each conversation to no more than one channel does not typically allow for fully exploiting the complete multilink available bandwidth. Typically, the residual bandwidth of each single link will be wasted because no conversation can fit into the remaining available capacity. Standard LAG does not allow to have links with different capacities and to change these capacities dynamically. Additionally, the standard does not allow links having two different capacities in the two directions. This asymmetrical situation may easily occur in presence of adaptive coded modulation. A non-exhaustive, brief description of some of the functionalities proposed by ML-PPP (RFC1 7 17, RFC2686) are given below: A session per link typically has to be established in order to exchange link capabilities {e.g. rate, format, compression,.. .). As a consequence of this negotiation mechanism, it is typically hard to apply a ML-PPP instance over links that change dynamically their own capacity, due to the adaptive coded modulation, and guarantee a hitless traffic distribution. The usage of adaptive coded modulation potentially may lead to asymmetrical link bandwidth due to different propagation conditions over the wireless media (typically the radio bit rate is symmetrical in both direction. However, as propagation phenomena may be different in forward and backward directions, this may lead to temporary situation in which bit rate becomes asymmetrical) Padding is contemplated. MLPPP foresees packets fragmentation and distribution over multiple links. Typically fragment are of the same size/length. Whenever a packets cannot be split in integer number of fragments, the last fragment is padded to reach the complete fragment size/length. Consequently, packet protection, namely traffic distribution without significantly disrupting the served traffic, is typically not available using the above technologies. Embodiments of the present invention feature a method of transmitting packets wherein said packets are comprised in a plurality of flows comprising frames, each flow comprising a corresponding plurality of flow characteristics comprising at least a committed information rate value and a class of service, the method comprising: - controlling admission of incoming flows; - mapping frames contained in admitted flows into fragments; - inserting a plurality of fragments having the same class of service into a queue; - obtaining a committed information rate value corresponding to said queue; - identifying a bandwidth available for transmission of said queue according to said obtained committed information rate value; - defining an order of transmission for the queue based on the class of service of the queue and the identified bandwidth; - generating a plurality of cells of the same size from a plurality of fragments; - distributing the plurality of cells between a plurality of individual transmission channels according to the order defined for transmission. According to some specific embodiments, a fragment has a determined size in terms of bytes and the step of mapping frames into fragments comprises: - mapping the entire frame in the fragment if the size of the frame is smaller than the determined size of the fragment; or - breaking the frame into portions of smaller size and mapping one or more of said portions in the fragment if the size of the frame is larger than the determined size of the fragment. According to some specific embodiments, one or more of said plurality of channels has a load prior to receiving one or more of said cells and said distribution of the plurality of cells on the plurality of individual transmission channels is made by loading a channel being less loaded prior to a channel being more loaded. According to some specific embodiments, the frames of an incoming flow comprise frame fields and said incoming flow is classified according to one or more frame fields of the frames comprised in said flow before being controlled for admission. According to some specific embodiments, the method is used in packet radio transmission. According to some specific embodiments, the method further comprises monitoring the usage of radio resources on each of said plurality of individual transmission channels thereby maintaining knowledge of the load present on a radio link used for such transmission. Some embodiments of the invention feature a transmitter for transmitting packets wherein said packets are comprised in a plurality of flows comprising frames, each flow comprising a corresponding plurality of flow characteristics comprising at least a committed information rate value and a class of service, the transmitter comprising: - an admission control module for controlling admission of incoming flows; - a compression and fragmentation module for mapping frames contained in admitted flows into fragments and for inserting a plurality of fragments having the same class of service into a queue; - a scheduler for obtaining a committed information rate value corresponding to said queue, for identifying a bandwidth available for transmission of said queue according to said obtained committed information rate value and for defining an order of transmission for the queue based on the class of service of the queue and the identified bandwidth; - a cell generator for generating a plurality of cells of the same size from a plurality of fragments; - a dispatcher for distributing the plurality of cells between a plurality of individual transmission channels according to the order defined for transmission. According to some specific embodiments, one or more of said plurality of channels has a load prior to receiving one or more of said cells and said dispatcher is configured for distributing said plurality of cells on the plurality of individual transmission channels by loading a channel being less loaded prior to a channel being more loaded. According to some specific embodiments the frames of an incoming flow comprise frame fields and the transmitter further comprises a classifier for classifying said incoming flow according to one or more frame fields of the frames comprised in said flow before being controlled for admission. According to some specific embodiments, the classifier is further configured to determine a type of compression and/or fragmentation and/or a class of service and/or a queue for inserting a fragment. According to some specific embodiments, the transmitter is configured to perform packet radio transmission. According to some specific embodiments, the dispatcher is configured for monitoring the usage of radio resources on each of said plurality of individual transmission channels thereby maintaining knowledge of the load present on a radio link used for such transmission. Some embodiments of the invention feature a receiver for receiving transmitted packets the receiver comprising: - a collector module for receiving a plurality of cells comprising payload information of flows, each cell comprising an identifier field for identifying an order of said cell; - a reordering module for reordering said received cells, in case said cells are received in disorder, according to the order of identifier field of each received cell; - a cell terminator and de-framer, for extracting the payload of each received cell, and generating a bit-stream from said payload and separating the bit-stream into fragments; - a plurality of de-compression modules, each configured reconstructing a frame from said fragments. According to some specific embodiments, the receiver is configured to receiver packets transmitted through radio transmission. Some embodiments of the invention feature a packet transmission and reception system comprising the transmitter and the receiver or a transceiver incorporating the transmitter and the receiver as proposed herein. Some embodiments of the invention feature a computer-executable or machineexecutable program product for the implementation of the steps of the method of transmission of packet as proposed herein when such program is run on a computer or a machine. These and further features and advantages of the present invention are described in more detail, for the purpose of illustration and not limitation, in the following description as well as in the claims with the aid of the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is an exemplary schematic representation of a conventional point-topoint multi-link radio transmission system. Figure 2 is an exemplary schematic representation of a point-to-point multi-link radio transmission system according to some embodiments. Figure 3 is an exemplary schematic representation of a transmission side of the point-to-point multi-link radio transmission system of figure 2. Figure 4 is an exemplary schematic representation of a reception side of the point-to-point multi-link radio transmission system of figure 2. DETAILED DESCRIPTION For a better understanding of the solution proposed herein, a brief description of an exemplary known system is first provided with reference to figure 1. In this figure, a conceptual scheme for one possible physical realization of a known point-to-point multi-link radio equipment is represented 1. As shown in the figure an incoming traffic 10 is input into an indoor unit 11 comprising a transmitter module 12 in charge of separating the incoming traffic into a plurality of independent radio channels 13. Each of said independent radio channels 13 is input into a respective microwave transmitter unit 14 which is in charge of transmitting an independent channel through a point-to-point radio connection as generally shown by reference numeral 16 through a branching module 15 which performs an analog sum of the signals to be transmitted. At the receive side, the individual channels are received by a branching module 25 which allows for discriminating and filtering the different radio channels within a received signal and subsequently forwarding the individual channels to respective microwave receiver unit 24. Each microwave receiver unit 24 at the receive side forwards a respective independent channel 23 towards a base-band processing unit 2 1 comprising a reception module 22 in charge of reconstructing the original traffic 20 and forward it to external equipment. The above description of a known point-to-point multi-link radio transmission system is only exemplary and other physical realizations are also possible. However, even if different configurations are used, the problem of an efficient use of the radio resources, as described above, when the input traffic is distributed over a plurality of individual radio channels is typically also present in such different configurations. Figure 2 illustrates an exemplary schematic representation of a point-to-point multi-link radio transmission system according to some embodiments. In figure 2, like elements have been given like reference numerals as those of figure 1. At the transmit side, a transmission mechanism receives a plurality of incoming flows for transmission through radio. The flows may be of different types {eg. Of different CoS) or sizes {eg. CIR, or bit rate in general) while radio channels may differ in channel space and/or throughput. It may be mentioned that according to embodiments proposed herein, the transmitter is configured so as to "consider", throughout the process for transmission, an array of multiple radio channels as a single virtual channel that may dynamically change its characteristics (for example throughput or delay). Such changes in the characteristics of the virtual channel may depend on radio propagation conditions or failure affecting one or more radio channels. In the context of the present specification, the term virtual channel refers to a compound of physical radio channels. Such channel is called virtual because it is not physical radio channel but it has substantially the same parameters of a physical radio channel such as available throughput and delay. Referring now to figure 2, an incoming traffic flow 10 is input into an indoor unit 11 comprising a processing equipment 17. The processing equipment 17 is in charge of processing the flow and provide such processed flow to output channels for transmission as will be described in further detail in relation to figure 3. A plurality of microwave transmitter units 14 are in charge of transmitting the processed flow through a point-to-point radio connection as generally shown by reference numeral 16 through a branching module 15. The microwave transmitter unit 14 may be an indoor or an outdoor unit. At the receive side, a plurality of cells comprising portions of flow are received by a branching module 25 and are input to respective microwave receiver units 24. Each microwave receiver unit 24 at the receive side forwards the cells towards a processing equipment 27 in charge of processing the received cells in order to reconstruct the original traffic 20 and forward it to external equipment as will be described in further detail in relation to figure 4 . The microwave receiver unit 24 may be an indoor or an outdoor unit. Conceptually, the transmission process may be summarized, for the sake of illustration and not limitation, in the preferred features described below: 1. Flow Admission Control: a dynamic admission control process is preferably executed in real-time depending on virtual channel characteristics. Such flow admission control procedure may admit some or all of the incoming flows in order to guarantee a desired Service Level Agreement as configured for a particular service. Such admission control may be performed according to the corresponding traffic descriptor which is information typically present in the flows and expressed in terms of for example committed information rate or peak information rate, type of service and CoS. 2. Compression and Fragmentation: a procedure capable of applying frame compression and/or fragmentation in order to achieve radio bandwidth optimization and reduce delay variation for frames belonging to a higher CoS (as these frames may be transmitted among fragments of packets with lower CoS). In such case, the fragment obtained from the original frame is preferably not padded to reach a minimum length. This is because the system is capable of operating with fragments of any suitable length. 3. Congestion management and consequent actions: when the available bandwidth of the virtual channel is not enough, as a result of a decrease in radio channel throughput and/or an increase in input flow rates, a mechanism is preferably employed which is able to queue packets and apply a process of discarding: a- frames corresponding to flows with input rate higher than the Committed one but less than the Peak Rate {i.e., the so-called "yellow" packets) corresponding to lower priority CoSs (starting from the lowest CoS first); or b- frames corresponding to flows with input rate within the Committed one {i.e., the so-called "green" packets) corresponding to lower priority CoSs (starting from the lowest CoS first); or c- entire flows with input rate within the Committed one, regardless of the CoSs the flows belong to, applicable only when current virtual channel available bandwidth is not enough to sustain such flows; or d- any combination of the above. The above process is preferably performed in the above order, a to c, however this is not mandatory and other orders may be applied. As mentioned above, a flow typically has a so-called traffic descriptor associated thereto which contains among others, two main parameters, one being the CIR that is the committed information rate a customer pays for to an operator and the other being the peak information rate (PIR) which is the maximum bit rate a customer is allowed to consume that is considered not guaranteed by the contract subscribed with the operator. Therefore, a bandwidth z between these two values, namely CIR