| NWCRG | V. Roca (Ed.) |
| Internet-Draft | INRIA |
| Intended status: Informational | J. Detchart |
| Expires: September 22, 2018 | ISAE - Supaero |
| C. Adjih | |
| INRIA | |
| M. Pedersen | |
| Steinwurf ApS | |
| I. Swett | |
| March 21, 2018 |
Generic Application Programming Interface (API) for Sliding Window FEC Codes
draft-roca-nwcrg-generic-fec-api-01
This document introduces a generic Application Programming Interface (API) for sliding window FEC codes. This API is meant to be compatible with any sliding window FEC code. It defines the core procedures and functions meant to control the codec (i.e., implementation of the FEC code), but leaves out all upper layer aspects that are the responsibility of the application or protocol making use of the codec. As a consequence, this is not an API for a FEC Scheme since certain mechanisms that must be defined by any FEC Scheme (e.g., signalling and FEC Payload IDs) are the responsibility of the caller instead of being addressed by the codec. A goal of this document is to pave the way for a future open-source implementation of such codes, another goal is to simplify the development of content delivery protocols that rely on sliding window FEC codes for robust transmissions.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 22, 2018.
Copyright (c) 2018 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Forward Erasure Correction (FEC) codes are a key element of communication systems, used to efficiently recover from packet losses during content delivery sessions. Among the FEC codes working at the network and higher layers, one can broadly distinguish block codes and sliding window codes. Block FEC codes require the data flow coming from the application to be segmented into blocks of a predefined maximum size, before generating a certain number of repair packets. With the second type of FEC codes, an encoding window continuously slides over the set of source data and repair packets are generated at any time by computing for instance a linear combination of data present in the encoding window. This fondamental difference seriously impacts the way they can be used by a content delivery protocol or application.
This document introduces a generic Application Programming Interface (API) for sliding window FEC codes. This API is meant to be usable by any sliding window FEC code and FEC Scheme independently of the protocol that may rely on it. This API defines the core procedures and functions meant to control the codec (i.e., implementation of the FEC code), but leaves out all upper layer aspects that are the responsibility of the application making use of the codec.
This API is meant to be usable by any sliding window FEC code. independently of the FEC Scheme or network coding protocol that may rely on it This API defines the core procedures and functions meant to control the codec (i.e., implementation of the FEC code), but leaves out all upper layer aspects that are the responsibility of the application making use of the codec. For instance, those restricted to end-to-end use-cases as well as those compatible with in-network re-encoding use-cases. Additionaly, this API is not impacted by the intra-flow versus inter-flow nature of the use-case, nor is it impacted by the single-path versus multi-paths nature of the use-case, since those are usage considerations under the responsibility of the caller.
A goal of this document is to pave the way for a future open-source implementation of such codes.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
This document uses the following definitions and abbreviations:
This generic FEC API is meant to be used with:
A key question is to determine what mechanisms are included in the codec and what mechanisms are left to the responsibility of the caller (i.e., an application or a protocol making use of this codec). More precisely, an FEC Scheme (such as the RLC FEC Scheme [RLC] in case of FECFRAME [fecframe-ext]) defines all the internal code details in order to enable interoperable implementations, but also signaling considerations that are essential to use them in a specific context.
Applying FEC in a given use-case to improve robustness involves many mechanisms. However they are not all the responsibility of the codec and can be instead implemented within the application or protocol that uses the codec. For instance the following mechanisms are considered out of scope of the API, being implemented by the caller, without any impact on the codec:
The following mechanisms are within scope of the API:
The following sections describe the generic API, following a C-language formalism. Everything is prefixed by XXX_, which stands for XXX.
<CODE BEGINS>
// all the definitions of interest
/**
* Function return value, indicating whether the function call succeeded or not.
*
* STATUS_OK = 0 Success
* STATUS_FAILURE, Failure. The function called did not succeed to perform
* its task, however this is not an error. This can happen
* for instance when decoding did not succeed (which is a
* valid output).
* STATUS_ERROR, Generic error type. The detailed error type is returned
* in a global variable, TBD_errno.
*/
typedef enum {
STATUS_OK = 0,
STATUS_FAILURE,
STATUS_ERROR
} TBD_status_t;
<CODE ENDS>
API proposal
<CODE BEGINS> <CODE ENDS>
Encoder API proposal
<CODE BEGINS> <CODE ENDS>
Decoder API proposal
TBD
N/A.
The authors would like to thank TBD.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [RFC6363] | Watson, M., Begen, A. and V. Roca, "Forward Error Correction (FEC) Framework", RFC 6363, DOI 10.17487/RFC6363, October 2011. |
| [fecframe-ext] | Roca, V. and A. Begen, "Forward Error Correction (FEC) Framework Extension to Sliding Window Codes", Transport Area Working Group (TSVWG) draft-ietf-tsvwg-fecframe-ext (Work in Progress), March 2018. |
| [RLC] | Roca, V., "Sliding Window Random Linear Code (RLC) Forward Erasure Correction (FEC) Scheme for FECFRAME", Transport Area Working Group (TSVWG) draft-ietf-tsvwg-rlc-fec-scheme (Work in Progress), March 2018. |
Editor's comment: we list a few existing APIs for reference and inspiration purposes. They will be removed in future versions of this document.
<CODE BEGINS>
// Copyright Steinwurf ApS 2011.
// Distributed under the "STEINWURF RESEARCH LICENSE 1.0".
// See accompanying file LICENSE.rst or
// http://www.steinwurf.com/licensing
class encoder
{
public:
/// Factory for encoders. The factory is used to build and initialize
/// encoders. If needed, e.g. for efficiency reasons, it is possible to
/// re-initialize already built encoders in order to reuse them.
class factory
{
public:
/// Constructor
factory();
/// @return The current specified symbol size in bytes.
uint64_t symbol_size() const;
/// @param symbol_size Sets the size of a symbol in bytes.
void set_symbol_size(uint64_t symbol_size);
/// @param field Set the finite field to use.
void set_field(finite_field field);
/// @return The finite field used.
finite_field field() const;
/// @return A new encoder.
encoder build();
/// @param encoder Initialize a encoder with the factory settings. After
/// calling initialize the encoder will be ready for use.
void initialize(encoder&);
[...]
};
public:
/// @return The total number of symbols available in memory at the encoder.
/// The number of symbols in the coding window MUST be less than
/// or equal to this number. The total range of valid symbol
/// indices is:
///
/// for (uint64_t i = 0; i < stream_symbols(); ++i)
/// {
/// std::cout << i + stream_lower_bound() << "\n";
/// }
///
uint64_t stream_symbols() const;
/// @return The index of the oldest symbol known by the encoder. This symbol
/// may not be inside the window but can be included in the window
/// if needed.
uint64_t stream_lower_bound() const;
/// @return The upper bound of the stream. The range of valid symbol indices
/// goes from [encoder::stream_lower_bound(),
/// encoder::stream_upper_bound()). Note the stream is a half-open
/// interval. Going from encoder::stream_lower_bound() to
/// encoder::stream_upper_bound() - 1.
uint64_t stream_upper_bound() const;
/// @return The size of a symbol in the stream in bytes.
uint64_t symbol_size() const;
/// Adds a new symbol to the front of the encoder. Increments the number of
/// symbols in the stream and increases the encoder::stream_upper_bound().
///
/// @param symbol Pointer to the symbol. Note, the caller must ensure that
/// the memory of the symbol remains valid as long as the symbol is
/// included in the stream. The caller is responsible for freeing the
/// memory if needed. Once the symbol is popped from the stream.
/// @return The stream index of the symbol being added.
uint64_t push_front_symbol(const uint8_t* symbol);
/// Remove the "oldest" symbol from the stream. Increments the
/// encoder::stream_lower_bound().
/// @return The index of the symbol being removed
uint64_t pop_back_symbol();
/// @return The number of symbols currently in the coding window. The
/// window must be within the bounds of the stream.
uint64_t window_symbols() const;
/// @return The index of the "oldest" symbol in the coding window.
uint64_t window_lower_bound() const;
/// @return The upper bound of the window. The range of valid symbol indices
/// goes from [encoder::window_lower_bound(),
/// encoder::window_upper_bound()). Note the window is a half-open
/// interval. Going from encoder::window_lower_bound() to
/// encoder::window_upper_bound() - 1.
uint64_t window_upper_bound() const;
/// The window represents the symbols which will be included in the next
/// encoding. The window cannot exceed the bounds of the stream.
///
/// Example: If window_lower_bound=4 and window_symbol=3 the following
/// symbol indices will be included 4,5,6
///
/// @param window_lower_bound Sets the index of the oldest symbol in the
/// window.
/// @param window_symbols Sets number of symbols within the window.
void set_window(uint64_t window_lower_bound, uint64_t window_symbols);
/// @return The size of the coefficient vector in the current window in
/// bytes. The number of coefficients is equal to the number of
/// symbols in the window. The size in bits of each coefficients
/// depends on the finite field chosen. A custom coding scheme can
/// be implemented by generating the coding vector manually.
/// Alternatively the built-in generator can be used. See
/// encoder::set_seed(...) and encoder::generate(...).
uint64_t coefficient_vector_size() const;
/// Seed the internal random generator function. If using the same seed
/// on the encoder and decoder the exact same set of coefficients will
/// be generated.
/// @param seed_value A value for the seed.
void set_seed(uint64_t seed_value);
/// Generate coding coefficients for the symbols in the coding window
/// according to the specified seed (see encoder::set_seed(...)).
/// @param coefficients Buffer where the coding coefficients should be
/// stored. This buffer must be encoder::coefficient_vector_size()
/// large in bytes.
void generate(uint8_t* coefficients);
/// Write an encoded symbol according to the coding coefficients.
/// @param symbol The buffer where the encoded symbol will be stored.
/// The symbol buffer must be encoder::symbol_size() large.
/// @param coefficients The coding coefficients. These must have the
/// memory layout required (see README.rst). A compatible format can
/// be created using encoder::generate(...)
void write_symbol(uint8_t* symbol, const uint8_t* coefficients);
/// Write a source symbol to the symbol buffer.
/// @param symbol The buffer where the source symbol will be stored. The
/// symbol buffer must be encoder::symbol_size() large.
/// @param index The symbol index which should be written.
void write_source_symbol(uint8_t* symbol, uint64_t index);
[...]
};
<CODE ENDS>
Morten API proposal
<CODE BEGINS>
// Copyright Steinwurf ApS 2011.
// Distributed under the "STEINWURF RESEARCH LICENSE 1.0".
// See accompanying file LICENSE.rst or
// http://www.steinwurf.com/licensing
class decoder
{
public:
class factory
{
public:
/// Constructor
factory();
/// @return The current specified symbol size in bytes.
uint64_t symbol_size() const;
/// @param symbol_size Sets the size of a symbol in bytes
void set_symbol_size(uint64_t symbol_size);
/// @param field Set the finite field to use
void set_field(finite_field field);
/// @return The finite field used.
finite_field field() const;
/// @return A new decoder.
decoder build();
/// @param decoder Initialize a decoder with the factory settings. After
/// calling initialize the decoder will be ready for use.
void initialize(decoder&);
[...]
};
public:
/// @return The total number of symbols known at the decoder. The number of
/// symbols in the decoding window MUST be less than or equal to
/// this number. The total range of valid symbol indicies is
///
/// for (uint64_t i = 0; i < stream_symbols(); ++i)
/// {
/// std::cout << i + stream_lower_bound() << "\n";
/// }
///
uint64_t stream_symbols() const;
/// @return The index of the oldest symbol known by the decoder. This symbol
/// may not be inside the window but can be included in the window
/// if needed.
uint64_t stream_lower_bound() const;
/// @return The upper bound of the stream. The range of valid symbol indices
/// goes from [decoder::stream_lower_bound(),
/// decoder::stream_upper_bound()). Note the stream is a half-open
/// interval. Going from decoder::stream_lower_bound() to
/// decoder::stream_upper_bound() - 1.
uint64_t stream_upper_bound() const;
/// @return The size of a symbol in the stream in bytes.
uint64_t symbol_size() const;
/// Adds a new symbol to the front of the decoder. Increments the number of
/// symbols in the stream and increases the decoder::stream_upper_bound().
///
/// @param symbol Pointer to the symbol. Note, the caller must ensure that
/// the memory of the symbol remains valid as long as the symbol is
/// included in the stream. The caller is responsible for freeing the
/// memory if needed. Once the symbol is popped from the stream.
/// @return The stream index of the symbol being added.
uint64_t push_front_symbol(uint8_t* symbol);
/// Remove the "oldest" symbol from the stream. Increments the
/// decoder::stream_lower_bound().
/// @return The index of the symbol being removed
uint64_t pop_back_symbol();
/// @return The number of symbols currently in the coding window. The
/// window must be within the bounds of the stream.
uint64_t window_symbols() const;
/// @return The index of the "oldest" symbol in the coding window.
uint64_t window_lower_bound() const;
/// @return The upper bound of the window. The range of valid symbol indices
/// goes from [decoder::window_lower_bound(),
/// decoder::window_upper_bound()). Note the window is a half-open
/// interval. Going from decoder::window_lower_bound() to
/// decoder::window_upper_bound() - 1.
uint64_t window_upper_bound() const;
/// The window represents the symbols which will be included in the next
/// decoding. The window cannot exceed the bounds of the stream.
///
/// Example: If window_lower_bound=4 and window_symbol=3 the following
/// symbol indices will be included 4,5,6
///
/// @param window_lower_bound Sets the index of the oldest symbol in the
/// window.
/// @param window_symbols Sets number of symbols within the window.
void set_window(uint64_t window_offset, uint64_t window_symbols);
/// @return The size of the coefficient vector in the current window in
/// bytes. The number of coefficients is equal to the number of
/// symbols in the window. The size in bits of each coefficients
/// depends on the finite field chosen. A custom coding scheme can
/// be implemented by generating the coding vector manually.
/// Alternatively the built-in generator can be used. See
/// decoder::set_seed(...) and decoder::generate(...).
uint64_t coefficient_vector_size() const;
/// Seed the internal random generator function. If using the same seed
/// on the decoder and decoder the exact same set of coefficients will
/// be generated.
/// @param seed_value A value for the seed.
void set_seed(uint64_t seed_value);
/// Generate coding coefficients for the symbols in the coding window
/// according to the specified seed (see decoder::set_seed(...)).
/// @param coefficients Buffer where the coding coefficients should be
/// stored. This buffer must be decoder::coefficient_vector_size()
/// large in bytes.
void generate(uint8_t* coefficients);
/// Decodes a coded symbol according to the coding coefficients.
///
/// Both buffers may be modified during this call. The reason for this
/// is that the decoder will directly operate on the provided memory
/// for performance reasons.
///
/// @param symbol Buffer representing a coded symbol.
///
/// @param coefficients The coding coefficients used to
/// create the encoded symbol
void read_symbol(uint8_t* symbol, uint8_t* coefficients);
/// Add a source symbol at the decoder.
///
/// @param symbol Buffer containing the source symbol's data.
/// @param index The index of the source symbol in the stream
void read_source_symbol(uint8_t* symbol, uint64_t index);
/// The rank of a decoder indicates how many symbols have been
/// partially or fully decoded. This number is also equivalent to the
/// number of pivot elements we have in the stream.
///
/// @return The rank of the decoder
uint64_t rank() const;
/// @return The number of missing symbols at the decoder
uint64_t symbols_missing() const;
/// @return The number of partially decoded symbols at the decoder
uint64_t symbols_partially_decoded() const;
/// @return The number of decoded symbols at the decoder
uint64_t symbols_decoded() const;
/// @param index Index of the symbol to check.
///
/// @return True if the symbol is decoded (i.e. it corresponds to a source
/// symbol), and otherwise false.
bool is_symbol_decoded(uint64_t index) const;
[...]
};
<CODE ENDS>
Morten API proposal
<CODE BEGINS>
/** a status for function calls **/
typedef enum {
STATUS_OK,
STATUS_ERROR,
/* ... */
} status_t;
/** defines the galois field used (at least the size, maybe we need to separate the implementations (Lookup Tables or Xor-based)) **/
typedef enum {
GF_16,
GF_64,
GF_256
} galois_field_t;
/*
* coding coefficient generators: specifies the algorithm used to generate the coefficients for the linear combinations
*
* NOTE: the choice of the finite field could done here (RLC_GF256, RLC_GF_16, ...) rather than using a different structure
*/
typedef enum {
RLC,
VDM
/* ... */
} coding_coefficients_generator_identifier_t;
/***
**
** encoder side
**
***/
/**
* NOTE for the callbacks in the sw_encoder: this is a proposition, we could also use 2 structures (source_t and repair_t) to avoid sending too many parameters in the callbacks.
**/
/**
* context: a context as a generic pointer (defined by the application if needed and given in sw_encoder_set_callbacks)
* src: the source data unit to consider
* src_id: the id of the source unit set by the encoder
* src_sz: the size in bytes of the data unit
**/
typedef void (*sw_encoder_callback_source_ready)(void *context,void* src, uint32_t src_id, size_t src_sz);
/**
* context: a context as a generic pointer (defined by the application if needed and given in sw_encoder_set_callbacks)
* rep: the repair data unit generated by the encoder
* rep_id: the id of the repair unit given by the encoder
* rep_sz: the size in bytes of the repair data unit
* src_ids: the array of the source unit ids used to generate the repair unit
* src_coefs: the coefficients used for each source units to generate the repair unit
* nb_src_in: number of src units in the linear combination (repair unit) (also the number of elements of src_ids and src_coefs)
**/
typedef void (*sw_encoder_callback_repair_ready)(void *context,void* rep, uint32_t rep_id, size_t rep_sz, uint32_t* src_ids, uint8_t *src_coefs, size_t nb_src_in);
/** a structure containing all we need to encode **/
typedef struct sw_encoder sw_encoder_t;
/**
* @brief Init a sliding window encoder by giving a galois field size, a coefficient generator function, and the maximum size of the window
* @param galois_field The size of the galois field used to create the coefficients and to encode.
* @param ccgi The function used to generate the coefficients (depends on the gf_size)
* @param max_wnd_size set the maximum of source units t consider inside the coding window (the oldest units will be destroyed)
* @return a sw_encoder_t structure.
**/
sw_encoder_t* sw_encoder_init(galois_field_t galois_field, coding_coefficients_generator_identifier_t ccgi, uint32_t max_wnd_size);
/**
* @brief Set the callbacks to get the encoded (repair) data
* @param encoder The encoder initialized
* @param context A generic context defined by the application (will be given in the callback)
* @param src_callback The function to be called when a source data unit has been processed by the encoder
* @param rep_callback The function to be called when a repair data unit has been generated
**/
status_t sw_encoder_set_callbacks(sw_encoder_t* encoder, void* context, sw_encoder_callback_source_ready src_callback, sw_encoder_callback_repair_ready rep_callback);
/**
* @brief Gives a source data unit and its id to the encoder
* @param encoder The encoder structure.
* @param src The data to add
* @param sz The size in bytes of the data unit
**/
status_t sw_encoder_add_source(sw_encoder_t* encoder, void* src, size_t sz);
/**
* @brief Removes the corresponding source data unit from the encoding window by giving and id
* @param encoder The encoder structure
* @param id The id of the data unit
* @return STATUS_OK if the data unit has been found and removed, STATUS_ERROR if the data unit doesn't exist
**/
status_t sw_encoder_remove_source(sw_encoder_t* encoder, uint32_t id);
/**
* @brief generates a repair data unit and calls the corresponding callback.
* @param encoder The encoder structure.
**/
status_t sw_encoder_generate_repair(sw_encoder_t* encoder);
/* ... */
status_t sw_encoder_set_control_parameter(sw_encoder_t* encoder, uint32_t type, void* value, uint32_t length);
/* ... */
status_t sw_encoder_get_control_parameter(sw_encoder_t* encoder, uint32_t type, void* value, uint32_t length);
/**
* @brief Release an encoder structure
* @param encoder The encoder structure
**/
status_t sw_encoder_release(sw_encoder_t* encoder);
/***
**
** decoder side
**
***/
/**
* NOTE for the callback in the sw_decoder: this is a proposition, we could also use an opaque structure (source_t) to avoid sending too many parameters in the callback.
**/
/**
* context: a context as a generic pointer (defined by the application if needed and given in sw_encoder_set_callbacks)
* src: the source data unit to consider
* src_id: the id of the source unit used in the decoder
* src_sz: the size in bytes of the data unit
**/
typedef void (*sw_decoder_callback_source_ready)(void *context,void* src, uint32_t src_id, size_t sz);
/** a structure containing all we need to decode **/
typedef struct sw_decoder sw_decoder_t;
/**
* @brief Init a sliding window decoder by giving a galois field size, a coefficient generator function, and the maximum size of the window
* @param galois_field The size of the galois field used to create the coefficients and to encode.
* @param ccgi The function used to generate the coefficients (depends on the gf_size)
* @return a sw_decoder_t structure.
**/
sw_decoder_t* sw_decoder_init(galois_field_t galois_field, coding_coefficients_generator_identifier_t ccgi);
/**
* @brief Set the callback to get the encoded (repair) data
* @param decoder The decoder initialized
* @param context A generic context defined by the application (will be given in the callback)
* @param callback The function to be called
**/
status_t sw_decoder_set_callback_source_ready(sw_decoder_t* decoder, void* context, sw_decoder_callback_source_ready callback);
/**
* NOTE for the next decode functions: we could also use 2 opaque structures to represent the source and repair units (source_t or repair_t)
**/
/**
* @brief decode some source data units by giving to the decoder a new source data unit
* @param decoder the decoder structure
* @param src the new source data unit
* @param src_id the id of the new source data unit (given by an encoder)
* @param src_sz the size in bytes of the new source data unit
**/
status_t sw_decoder_decode_with_source(sw_decoder_t* decoder, void* src, uint32_t src_id, size_t src_sz);
/**
* @brief decode some source data units by giving to the decoder a new repair data unit
* @param decoder the decoder structure
* @param rep the new repair data unit
* @param rep_id the id of the new repair data unit (given by an encoder)
* @param rep_sz the size in bytes of the new repair data unit
* @param src_ids: the array of the source unit ids used to generate this repair unit
* @param src_coefs: the coefficients used for each source units to generate this repair unit
* @param nb_src_in: number of src units in the linear combination (repair unit) (also the number of elements of src_ids and src_coefs)
**/
status_t sw_decoder_decode_with_repair(sw_decoder_t* decoder, void* rep, uint32_t rep_id, size_t rep_sz, uint32_t* src_ids, uint8_t * src_coefs, size_t nb_src_in);
/* ... */
status_t sw_decoder_set_control_parameter(sw_decoder_t* decoder, uint32_t type, void* value, uint32_t length);
/* ... */
status_t sw_decoder_get_control_parameter(sw_decoder_t* decoder, uint32_t type, void* value, uint32_t length);
/**
* @brief Release a decoder structure
* @param decoder The decoder structure to release
**/
status_t sw_decoder_release(sw_decoder_t* decoder);
<CODE ENDS>
Jonathan API proposal
For DRAGONCAST/DragonNet/GardiNet (<https://gitlab.inria.fr/GardiNet/liblc/>):
<CODE BEGINS>
/**
* The fec_codec_id_t enum identifies the FEC code/codec being used.
* Since a given fec_codec_id can be used by one or several FEC schemes (that specify
* both the codes and way of using these codes), it is distinct from the FEC Encoding
* ID.
*/
typedef enum {
CODEC_NIL = 0,
CODEC_RLC
} codec_id_t;
/**
* Function return value, indicating whether the function call succeeded or not.
* In case of failure, the detailed error type is returned in a global variable,
* of_errno (see of_errno.h).
*
* STATUS_OK = 0 Success
* STATUS_FAILURE, Failure. The function called did not succeed to perform
* its task, however this is not an error. This can happen
* for instance when decoding did not succeed (which is a
* valid output).
* STATUS_ERROR, Generic error type. The caller is expected to be able
* to call the library in the future after having corrected
* the error cause.
* STATUS_FATAL_ERROR Fatal error. The caller is expected to stop using this
* codec instance immediately (it replaces an exit() system
* call).
*/
typedef enum {
STATUS_OK = 0,
STATUS_FAILURE,
STATUS_ERROR,
STATUS_FATAL_ERROR
} status_t;
/**
* Throughout the API, a pointer to this structure is used as an identifier of the current
* codec instance, also known as "session".
*
* This generic structure is meant to be extended by each codec and new pieces of information
* that are specific to each codec be specified there. However, all the codec specific structures
* MUST begin the same entries as the ones provided in this generic structure, otherwise
* hazardous behaviors may happen.
*/
typedef struct session {
codec_id_t codec_id;
codec_type_t codec_type;
} session_t;
/**
* Generic FEC parameter structure used by set_fec_parameters().
*
* This generic structure is meant to be extended by each codec and new pieces of information
* that are specific to each codec be specified there. However, all the codec specific structures
* MUST begin the same entries as the ones provided in this generic structure, otherwise
* hazardous behaviors may happen.
*/
typedef struct {
/** SENDER and RECEIVER: maximum number of source symbols used for any repair symbol. */
UINT32 coding_window_max_size;
/** RECEIVER only: maximum number of source symbols kept in current linear
* system. If the linear system grows above this limit, older source symbols
* in excess are removed and the application callback called if set. This
* value MUST be larger than the coding_window_max_size. */
UINT32 linear_system_max_size;
UINT32 encoding_symbol_length;
} parameters_t;
<CODE ENDS>
Vincent API proposal
<CODE BEGINS>
/**
* This function allocates and partially initializes a new session structure.
* Throughout the API, a pointer to this session is used as an identifier of the
* current codec instance.
*
* @param ses (IN/OUT) address of the pointer to a session. This pointer is updated
* by this function.
* In case of success, it points to a session structure allocated by the
* library. In case of failure it points to NULL.
* @param codec_id identifies the FEC code/codec being used.
* @param codec_type indicates if this is a coder or a decoder.
* @param verbosity set the verbosity level
* @return Completion status. The ses pointer is updated according to the success return status.
*/
status_t create_codec_instance (session_t** ses,
codec_id_t codec_id,
codec_type_t codec_type,
uint32_t verbosity);
/**
* This function releases all the internal resources used by this FEC codec instance.
* None of the source symbol buffers will be free'ed by this function, even those decoded by
* the library if any, regardless of whether a callback has been registered or not. It's the
* responsibility of the caller to free them.
*
* @param ses (IN) Pointer to the session.
* @return Completion status
*/
status_t release_codec_instance (session_t* ses);
/**
* Second step of the initialization, where the application specifies code(c) specific parameters.
*
* At a receiver, the parameters can be extracted from the FEC OTI that is usually communicated
* to the receiver by either an in-band mechanism or an out-of-band mechanism, or set statically
* for a specific use-case.
*
* @param ses (IN) Pointer to the session.
* @param params (IN) pointer to a structure containing the FEC parameters associated to
* a specific FEC codec.
* @return Completion status.
*/
status_t set_fec_parameters (session_t* ses,
parameters_t* params);
/**
* This function sets a FEC scheme/FEC codec specific control parameter,
* using a type/value method.
*
* @param ses (IN) Pointer to the session.
* @param type (IN) Type of parameter. This type is FEC codec ID specific.
* @param value (IN) Pointer to the value of the parameter. The type of the object pointed
* is FEC codec ID specific.
* @param length (IN) length of pointer value
* @return Completion status.
*/
status_t set_control_parameter (session_t* ses,
UINT32 type,
void* value,
UINT32 length);
/**
* This function gets a FEC scheme/FEC codec specific control parameter,
* using a type/value/length method.
*
* @param ses (IN) Pointer to the session.
* @param type (IN) Type of parameter. This type is FEC codec ID specific.
* @param value (IN/OUT) Pointer to the value of the parameter. The type of the object
* pointed is FEC codec ID specific. This function updates the value object
* accordingly. The application, who knows the FEC codec ID, is responsible
* to allocating the appropriate object pointed by the value pointer.
* @param length (IN) length of pointer value
* @return Completion status.
*/
status_t get_control_parameter (session_t* ses,
UINT32 type,
void* value,
UINT32 length);
<CODE ENDS>
Vincent API proposal
<CODE BEGINS>
/**
* Set the various callback functions for this session.
* All the callback functions require an opaque context parameter, that must be
* initialized accordingly by the application, since it is application specific.
*
* @param ses (IN) Pointer to the session.
*
* @param decoded_source_symbol_callback
* (IN) Pointer to the function, within the application, that
* needs to be called each time a source symbol is decoded.
*
* @param available_source_symbol_callback
* (IN) Pointer to the function, within the application, that
* needs to be called each time a source symbol is decoded and
* all computations performed (i.e., the buffer does contain the
* symbol value).
*
* @param source_symbol_removed_from_coding_window_callback
* (IN) Pointer to the function, within the application, that
* needs to be called each time a source symbol is removed from
* the left side of the coding window, at a SENDER because this
* window has slided to the right, or at a RECEIVER because this
* old source symbol is now forgotten.
*
* @param context_4_callback
* (IN) Pointer to the application-specific context that will be
* passed to the callback function (if any). This context is not
* interpreted by this function.
*
* @return Completion status.
*/
status_t set_callback_functions (of_session_t* ses,
void* (*decoded_source_symbol_callback) (void *context,
UINT32 size, /* size of decoded source symbol */
UINT32 esi), /* encoding symbol ID */
void (*available_source_symbol_callback) (void *context,
void *new_symbol_buf,/* symbol buffer */
UINT32 size, /* size of decoded source symbol */
UINT32 esi), /* encoding symbol ID */
void (*source_symbol_removed_from_coding_window_callback)
(void *context,
UINT32 old_symbol_esi),
void* context_4_callback);
<CODE ENDS>
Vincent API proposal
TBD
<CODE BEGINS>
/**
* SENDER: this function specifies the coding coefficients chosen by the application if this is the way the codec
* works. This function MUST be called before calling build_repair_symbol().
* RECEIVER: communicate the coding coefficients associated to a repair symbol and carried in the packet header.
* This function MUST be called before calling decode_with_new_repair_symbol().
*
* @param ses
* @param coding_coefs_tab (IN) table of coding coefficients to be associated to each of the source symbols
* currently in the coding window. The size (number of bits) of each coefficient
* depends on the FEC scheme. The allocation and release of this table is under the
* responsibility of the application.
* @param nb_coefs_in_tab (IN) number of entries (i.e., coefficients) in the table.
* @return Completion status.
*/
status_t set_coding_coefficients_tab (session_t* ses,
void* coding_coefs_tab,
UINT32 nb_coefs_in_tab);
/**
* SENDER: this function enables the application to retrieve the set of coding coefficients generated and used by
* build_repair_symbol().
* RECEIVER: never used.
*
* @param ses
* @param coding_coefs_tab (IN/OUT) pointer of a table of coding coefficients to be associated to each of the
* source symbols currently in the coding window. The size (number of bits) of each
* coefficient depends on the FEC scheme. The allocation and release of this table is
* under the responsibility of the application. Upon return of this function, this
* table is allocated and filled with each coefficient value.
* @param nb_coefs_in_tab (IN/OUT) pointer to the number of entries (i.e., coefficients) in the table.
* Upon calling this function, this number must be zero. Upon return of this function
* this number is initialized with the actual number of entries in the coeffs_tab[].
* @return Completion status (OF_STATUS_OK, FAILURE, ERROR or FATAL_ERROR).
*/
status_t get_coding_coefficients_tab (session_t* ses,
void** coding_coefs_tab,
UINT32* nb_coefs_in_tab);
/**
* The coding coefficients may be generated in a deterministic manner, (e.g., through the use of a PRNG and the
* repair symbol ESI used as a seed). This is the case with RLC codes.
*
* SENDER: generate all coefficients. This function MUST be called before calling build_repair_symbol().
* RECEIVER: generate all coefficients. This function MUST be called before calling decode_with_new_repair_symbol().
*
* @param ses
* @param params (IN) pointer to a codec specific structure containing the required parameters.
* These parameters can include a repair symbol ESI or key among other things.
* @return Completion status.
*/
status_t generate_coding_coefficients (session_t* ses,
void* params);
<CODE ENDS>
Vincent API proposal
<CODE BEGINS>
/**
* Create a single repair symbol, i.e. perform an encoding.
* This function requires that the application has previously set the coding window and if needed the coding coefficients
* appropriately. After that, the application can call this function.
*
* @param ses
* @param new_repair_symbol_buf (IN) The pointer to the buffer for the repair symbol to build can either point to a buffer
* allocated by the application, or let to NULL meaning that this function will allocate
* memory.
* @return Completion status.
*/
status_t ccod_build_repair_symbol (session_t* ses,
void* new_repair_symbol_buf);
<CODE ENDS>
Vincent API proposal
<CODE BEGINS>
/**
* Submit a received source symbol and try to progress in the decoding. For each decoded source
* symbol, if any, the application is informed through the dedicated callback functions.
*
* This function usually returns OF_STATUS_OK, regardless of whether this new symbol enabled the
* decoding of one or several source symbols, unless an error occured. This function cannot return
* OF_STATUS_FAILURE.
*
* @param ses
* @param new_src_symbol_buf (IN) Pointer to the new source symbol now available (i.e. a new symbol received by
* the application, or a decoded symbol in case of a recursive call if it makes sense).
* @param new_symbol_esi_or_key (IN) encoding symbol ID of the new source symbol or key if there is no notion of ESI.
* @return Completion status.
*/
status_t decode_with_new_source_symbol (session_t* ses,
void* const new_src_symbol_buf,
UINT32 new_symbol_esi_or_key);
/**
* Submit a received repair symbol and try to progress in the decoding. For each decoded source
* symbol, if any, the application is informed through the dedicated callback functions.
*
* This function requires that the application has previously set the coding window and the coding coefficients appropriately.
* After that, the application can call this function. The
* application keeps a full control of the repair symbol buffer, i.e., the application is in charge
* of freeing this buffer as soon as it believes appropriate to do so (a copy is kept by the codec).
*
* This function usually returns OF_STATUS_OK, regardless of whether this new symbol enabled the
* decoding of one or several source symbols, unless an error occurred. This function cannot return
* OF_STATUS_FAILURE.
*
* @param ses
* @param new_repair_symbol_buf (IN) Pointer to the new repair symbol now available (i.e. a new symbol received by
* the application or a decoded symbol in case of a recursive call if it makes sense).
* @return Completion status.
*/
status_t decode_with_new_repair_symbol (session_t* ses,
void* const new_repair_symbol_buf);
<CODE ENDS>
Vincent API proposal