File Protocol¶
The file protocol is implemented by both the file transfer service and clients interacting with the service. This protocol uses a content-addressable methodology similar to git for storing and chunking files. This document covers the content-addressable storage, all messages used in the protocol, and includes diagrams of common use cases.
Content-Addressable Storage¶
The file protocol uses a content-addressable system to store file data.
All files are broken up into 4KB chunks prior to sending. This chunking
is initiated either by an export or import message. A local
folder called storage is created by the file service and client
for storing the content-addressable information. Inside of storage
each file has it’s own folder. The folder name is a 16-bit
blake2 hash of the file’s
contents. This folder is created as part of the import/export process.
Inside of each file’s folder there is a meta file and chunk files.
Each meta file contains metadata about the file in a CBOR list.
Currently we only store the number of chunks in the meta file.
Each chunk file is also stored in this folder. Chunk files are named
after the hex representation of their chunk number. Each chunk file
contains the raw contents of that chunk.
Here is an example content-addressable storage structure containing an eleven chunk file:
storage/
└── 852f1630f4ed2c0bc934d71ada618974/ <- blake2 hash of file
├── 0 <- Each of these are file chunks
├── 1
├── 2
├── 3
├── 4
├── 5
├── 6
├── 7
├── 8
├── 9
├── A
└── meta <- Contains {"num_chunks" : 11 } in CBOR
Messages¶
All messages in the file protocol are encoded as CBOR arrays and are sent
in UDP packets. The first value in the encoded list is the channel_id
for request/response type messages and the hash for content-addressable
messages.
The hash parameter is the blake2 hash for the corresponding file
which is being transferred.
The channel_id parameter corresponds to an in-memory array of coroutines
created as part of the file transfer process.
Send Chunk¶
This message is sent as part of the file import or export process.
It contains the file hash, chunk index and raw chunk data.
Each raw chunk is 4KB in size. Individual chunk messages will not get
an immediate reply. However if no chunks are received within the
timeout window then an ack or nak will be sent depending
on whether all the chunks have been received or not.
{ hash, chunk_index, data }
Sync¶
This message is sent to query the message receiver on the status of a file. It contains the file’s hash and the expected number of chunks for the file.
If the file does not exist on the receiver’s side then the receiver
will send a nak requesting all chunks and create the appropriate
file storage structure. If the file does exist then the receiver will
send back an ack. The num_chunks is optional, however it
should be sent in the first sync of an import or export
to ensure the expected number of chunks is known.
{ hash, num_chunks }
Acknowledge (Ack)¶
This message is sent to inform the message receiver that the message sender has all chunks for a given file. It contains the file’s hash, the boolean value true, and the number of chunks in the file.
{ hash, true, num_chunks }
Negative Acknowledge (Nak)¶
This message is sent to inform the message receiver that the
message sender does not have all chunks for a given file. It
contains the file’s hash, the boolean value false, and a list
of missing chunks. The list of missing chunks is made up of
pairs of ranges where the first number is inclusive and the
second is exclusive. For example 0, 2 means the first
two chunks are missing.
A nak may be sent in response to a sync or after a
timeout during a file import or export. The message
sender should expect the message receiver to send
the missing file chunks upon receipt of a nak.
{ hash, false 1, 4, 6, 7 }
The above example nak indicates that chunks 1-3 and 6
are missing.
Export¶
This message is sent to initiate the process of transferring a file from the message sender to the message receiver. It contains the channel id, the string “export”, the file’s hash, the target path for the file and file’s permissions mode.
The message receiver will begin waiting for file chunks after
receiving this message. Once the timeout triggers it will
attempt to export the file locally. If the file is incomplete then
the receiver will request any missing chunks. Upon receiving
all chunks it will attempt to verify and export the file to
the local filesystem. This message is sent after the
sync command as part of the export process.
{ channel_id, "export", hash, path, mode }
Import¶
This message is sent to initiate the process of transferring a file to the message sender from the message receiver. It contains the channel ID, the string “import”, and the requested file’s path.
Upon receiving, the message receiver will import the requested
file into the managed content-addressable storage and send a
success message to the sender. This success message
will contain the file`s hash and allow the original message
sender to determine which file chunks are required.
{ channel_id, "import", path }
Success¶
This message is sent as part of the import or export
processes. It contains the channel ID, the boolean value true
and potentially other values depending on the situation.
This message is primarily sent in two different situations:
at the end of an export and near the beginning of an import.
The message sender would send a success if an export
has completed successfully. The success is also used
during an import to indicate a file is ready for sending
and to communicate the file’s hash.
Extra values in this command appear as extra items in the list.
{ channel_id, true, ..values }
Failure¶
This message is sent if there as an error in the import or
export process. It contains the channel ID, the boolean false
and the error message.
{ channel_id, false, error_message }
Common Protocol Usages¶
Uploading a single chunk file from a ground station to an OBC:
Downloading a single chunk file from an OBC to a ground station:
Uploading a three chunk file from ground station with a chunk re-request:
