In the IRI Node
class, this code segment checks to see if the requested transaction hash is the same as the received transaction hash:
https://github.com/iotaledger/iri/blob/dev/src/main/java/com/iota/iri/network/Node.java#L284
//Request bytes
//add request to reply queue (requestedHash, neighbor)
Hash requestedHash = new Hash(receivedData, TransactionViewModel.SIZE, reqHashSize);
if (requestedHash.equals(receivedTransactionHash)) {
//requesting a random tip
requestedHash = Hash.NULL_HASH;
}
How is it possible for the request hash to be part of the received transaction hash since the received transaction hash is not known until the transaction is made?
Perhaps I am missing some detail - can someone please explain the purpose of this.
History of This Code Segment
After the first submitted answer below, and after Come_From_Beyond made some comments, I did some digging to see the commit history of this code.
According to the earliest available commits for the Node
class, when the transaction hash and the requested hash were the same, the request was to be the Milestone that is requested. Perhaps someone can shed some light on why this existed like this and why it was changed.
Oct 24, 2016 - If the requested hash is same, then return a Milestone https://github.com/iotaledger/iri/blob/5db789222fabd11a9880b457b2b9f24cd783789f/src/iri/Node.java#L90
if (Arrays.equals(requestedTransaction, receivedTransaction.hash)) {
transactionPointer = Storage.transactionPointer(Milestone.latestMilestone.bytes);
} else {
transactionPointer = Storage.transactionPointer(requestedTransaction);
}
Dec 23, 2016 - Introduces Concept of 'Experimental Random Tip Broadcaster' https://github.com/iotaledger/iri/blob/577f2ad2db561a4044cbf9a81beae15da599d2ab/src/main/java/com/iota/iri/service/Node.java#L201
if (Arrays.equals(requestedTransaction, receivedTransaction.hash)) {
if (Configuration.booling(DefaultConfSettings.EXPERIMENTAL) &&
++randomTipBroadcastCounter % 3 == 0) {
// see code re "Experimental: Random Tip Broadcaster"
} else {
transactionPointer = StorageTransactions.instance()
.transactionPointer(Milestone.latestMilestone.bytes());
}
} else {
transactionPointer = StorageTransactions.instance().transactionPointer(requestedTransaction);
}
Jan 29, 2017 - Compare changed to NULL HASH and other changes to scheme https://github.com/iotaledger/iri/blob/61ad77be4ce01ee1cc8530320f7c234ee660c822/src/main/java/com/iota/iri/service/Node.java#L205
(truncated for brevity)
if (Arrays.equals(requestedTransaction, Hash.NULL_HASH.bytes())) {
if (randomTipBroadcastCounter % 9 == 0) {
// see code "Random Broadcaster - latestMilestone index==0."
}
else if (randomTipBroadcastCounter % 6 == 0) {
// see code re Milestone
}
else if (randomTipBroadcastCounter % 3 == 0) {
// see code "Random Broadcaster - random tip."
}
} else {
transactionPointer = StorageTransactions.instance().transactionPointer(requestedTransaction);
}
The Role of P_MILESTONE TO SELECT
and the TransactionRequester
The actual actions of the transaction to request are further obfuscated by the last minute code block before transmission in Node
.
See https://github.com/iotaledger/iri/blob/dev/src/main/java/com/iota/iri/network/Node.java#L483
synchronized (sendingPacket) {
System.arraycopy(transactionViewModel.getBytes(), 0, sendingPacket.getData(), 0, TransactionViewModel.SIZE);
Hash hash = transactionRequester.transactionToRequest(rnd.nextDouble() < P_SELECT_MILESTONE);
System.arraycopy(hash != null ? hash.bytes() : transactionViewModel.getHash().bytes(), 0,
sendingPacket.getData(), TransactionViewModel.SIZE, reqHashSize);
neighbor.send(sendingPacket);
}
By default, P_MILESTONE_TO_SELECT
is set to 70%, and this tells the transactionRequester
to use a Milestone if any are found in the milestoneTransactionsToRequest
queue.
This further obfuscates the actions of a request, although a little investigation will find that during validation, if any trunk or branch is found to be a Milestone, then it is added to transactionsToRequest
. And also, if the LedgerValidator thread also finds any missing Milestones, they too are added to the transactionsToRequest
queue.
This complicates the logic of a requested transaction quite a bit as it turns out that a requested transaction of either (1) an identical hash or (2) NULL_HASH, has the effect of not //requesting a random tip
as the comments in Node
imply, but of an elaborate process of preferring a Milestone if one is missing from the system.
From what I can tell of the effect of the LedgerValidator
, all tips are checked during the routine Milestone.java
checking thread Solid Milestone Tracker
and when the LedgerValidator
is called, missing Milestones are inventoried and queued to receive priority treatment within the system.
I suppose this begs the question ...
How often is a Milestone missing from the system? That determines, ultimately, what a requested transaction actually requests.