The previous sections have addressed two out of three questions posed in the introduction: “what the ledger looks like”, and “who may request which changes”. This section addresses the last one, “who sees which changes and data”. That is, it explains the privacy model for Daml ledgers.

The privacy model of Daml Ledgers is based on a need-to-know basis, and provides privacy on the level of subtransactions. Namely, a party learns only those parts of ledger changes that affect contracts in which the party has a stake, and the consequences of those changes. And maintainers see all changes to the contract keys they maintain.

To make this more precise, a stakeholder concept is needed.

Contract Observers and Stakeholders

Intuitively, as signatories are bound by a contract, they have a stake in it. Actors might not be bound by the contract, but they still have a stake in their actions, as these are the actor’s rights. Generalizing this, observers are parties who might not be bound by the contract, but still have the right to see the contract. For example, Alice should be an observer of the PaintOffer, such that she is made aware that the offer exists.

Signatories are already determined by the contract model discussed so far. The full contract model additionally specifies the contract observers on each contract. A stakeholder of a contract (according to a given contract model) is then either a signatory or a contract observer on the contract. Note that in Daml, as detailed later, controllers specified using simple syntax are automatically made contract observers whenever possible.

In the graphical representation of the paint offer acceptance below, contract observers who are not signatories are indicated by an underline.

The paint offer acceptance flowchart. In the first subtransaction and in EXE A (IOU Bank A), A is underlined; in IOU $Bank P, P is underlined.

Choice Observers

In addition to contract observers, the contract model can also specify choice observers on individual Exercise actions. Choice observers get to see a specific exercise on a contract, and to view its consequences. Choice observers are not considered stakeholders of the contract, they only affect the set of informees on an action, for the purposes of projection (see below).


Stakeholders should see changes to contracts they hold a stake in, but that does not mean that they have to see the entirety of any transaction that their contract is involved in. This is made precise through projections of a transaction, which define the view that each party gets on a transaction. Intuitively, given a transaction within a commit, a party will see only the subtransaction consisting of all actions on contracts where the party is a stakeholder. Thus, privacy is obtained on the subtransaction level.

An example is given below. The transaction that consists only of Alice’s acceptance of the PaintOffer is projected for each of the three parties in the example: the painter, Alice, and the bank.

The original paint offer flowchart followed by two projections, one for P and A and one for the bank. The privacy implications of what is visible in each projection are detailed in the next three paragraphs.

Since both the painter and Alice are stakeholders of the PaintOffer contract, the exercise on this contract is kept in the projection of both parties. Recall that consequences of an exercise action are a part of the action. Thus, both parties also see the exercise on the Iou Bank A contract, and the creations of the Iou Bank P and PaintAgree contracts.

The bank is not a stakeholder on the PaintOffer contract (even though it is mentioned in the contract). Thus, the projection for the bank is obtained by projecting the consequences of the exercise on the PaintOffer. The bank is a stakeholder in the contract Iou Bank A, so the exercise on this contract is kept in the bank’s projection. Lastly, as the bank is not a stakeholder of the PaintAgree contract, the corresponding Create action is dropped from the bank’s projection.

Note the privacy implications of the bank’s projection. While the bank learns that a transfer has occurred from A to P, the bank does not learn anything about why the transfer occurred. In practice, this means that the bank does not learn what A is paying for, providing privacy to A and P with respect to the bank.

As a design choice, Daml Ledgers show to contract observers only the state changing actions on the contract. More precisely, Fetch and non-consuming Exercise actions are not shown to contract observers - except when they are also actors or choice observers of these actions. This motivates the following definition: a party p is an informee of an action A if one of the following holds:

  • A is a Create on a contract c and p is a stakeholder of c.
  • A is a consuming Exercise on a contract c, and p is a stakeholder of c or an actor on A. Note that a Daml choice controller can be an exercise actor without being a contract stakeholder.
  • A is a non-consuming Exercise on a contract c, and p is a signatory of c or an actor on A.
  • A is an Exercise action and p is a choice observer on A.
  • A is a Fetch on a contract c, and p is a signatory of c or an actor on A.
  • A is a NoSuchKey k assertion and p is a maintainer of k.

Then, we can formally define the projection of a transaction tx = act1, …, actn for a party p is the subtransaction obtained by doing the following for each action acti:

  1. If p is an informee of acti, keep acti as-is.
  2. Else, if acti has consequences, replace acti by the projection (for p) of its consequences, which might be empty.
  3. Else, drop acti.

Finally, the projection of a ledger l for a party p is a list of transactions obtained by first projecting the transaction of each commit in l for p, and then removing all empty transactions from the result. Note that the projection of a ledger is not a ledger, but a list of transactions. Projecting the ledger of our complete paint offer example yields the following projections for each party:

Time sequences for each party's projection, explained in-depth in the numbered list below.

Examine each party’s projection in turn:

  1. The painter does not see any part of the first commit, as he is not a stakeholder of the Iou Bank A contract. Thus, this transaction is not present in the projection for the painter at all. However, the painter is a stakeholder in the PaintOffer, so he sees both the creation and the exercise of this contract (again, recall that all consequences of an exercise action are a part of the action itself).
  2. Alice is a stakeholder in both the Iou Bank A and PaintOffer A B Bank contracts. As all top-level actions in the ledger are performed on one of these two contracts, Alice’s projection includes all the transactions from the ledger intact.
  3. The Bank is only a stakeholder of the IOU contracts. Thus, the bank sees the first commit’s transaction as-is. The second commit’s transaction is, however dropped from the bank’s projection. The projection of the last commit’s transaction is as described above.

Ledger projections do not always satisfy the definition of consistency, even if the ledger does. For example, in P’s view, Iou Bank A is exercised without ever being created, and thus without being made active. Furthermore, projections can in general be non-conformant. However, the projection for a party p is always

  • internally consistent for all contracts,
  • consistent for all contracts on which p is a stakeholder, and
  • consistent for the keys that p is a maintainer of.

In other words, p is never a stakeholder on any input contracts of its projection. Furthermore, if the contract model is subaction-closed, which means that for every action act in the model, all subactions of act are also in the model, then the projection is guaranteed to be conformant. As we will see shortly, Daml-based contract models are conformant. Lastly, as projections carry no information about the requesters, we cannot talk about authorization on the level of projections.

Privacy Through Authorization

Setting the maintainers as required authorizers for a NoSuchKey assertion ensures that parties cannot learn about the existence of a contract without having a right to know about their existence. So we use authorization to impose access controls that ensure confidentiality about the existence of contracts. For example, suppose now that for a PaintAgreement contract, both signatories are key maintainers, not only the painter. That is, we consider PaintAgreement @A @P &P123 instead of PaintAgreement $A @P &P123. Then, when the painter’s competitor Q passes by A’s house and sees that the house desperately needs painting, Q would like to know whether there is any point in spending marketing efforts and making a paint offer to A. Without key authorization, Q could test whether a ledger implementation accepts the action NoSuchKey (A, P, refNo) for different guesses of the reference number refNo. In particular, if the ledger does not accept the transaction for some refNo, then Q knows that P has some business with A and his chances of A accepting his offer are lower. Key authorization prevents this flow of information because the ledger always rejects Q‘s action for violating the authorization rules.

For these access controls, it suffices if one maintainer authorizes a NoSuchKey assertion. However, we demand that all maintainers must authorize it. This is to prevent spam in the projection of the maintainers. If only one maintainer sufficed to authorize a key assertion, then a valid ledger could contain NoSuchKey k assertions where the maintainers of k include, apart from the requester, arbitrary other parties. Unlike Create actions to contract observers, such assertions are of no value to the other parties. Since processing such assertions may be expensive, they can be considered spam. Requiring all maintainers to authorize a NoSuchKey assertion avoids the problem.

Divulgence: When Non-Stakeholders See Contracts

The guiding principle for the privacy model of Daml ledgers is that contracts should only be shown to their stakeholders. However, ledger projections can cause contracts to become visible to other parties as well.

In the example of ledger projections of the paint offer, the exercise on the PaintOffer is visible to both the painter and Alice. As a consequence, the exercise on the Iou Bank A is visible to the painter, and the creation of Iou Bank P is visible to Alice. As actions also contain the contracts they act on, Iou Bank A was thus shown to the painter and Iou Bank P was shown to Alice.

Showing contracts to non-stakeholders through ledger projections is called divulgence. Divulgence is a deliberate choice in the design of Daml ledgers. In the paint offer example, the only proper way to accept the offer is to transfer the money from Alice to the painter. Conceptually, at the instant where the offer is accepted, its stakeholders also gain a temporary stake in the actions on the two Iou contracts, even though they are never recorded as stakeholders in the contract model. Thus, they are allowed to see these actions through the projections.

More precisely, every action act on c is shown to all informees of all ancestor actions of act. These informees are called the witnesses of act. If one of the witnesses W is not a stakeholder on c, then act and c are said to be divulged to W. Note that only Exercise actions can be ancestors of other actions.

Divulgence can be used to enable delegation. For example, consider the scenario where Alice makes a counteroffer to the painter. Painter’s acceptance entails transferring the IOU to him. To be able to construct the acceptance transaction, the painter first needs to learn about the details of the IOU that will be transferred to him. To give him these details, Alice can fetch the IOU in a context visible to the painter:

A series of time sequences showing how delegation changes privacy, as described in the preceding paragraph.

In the example, the context is provided by consuming a ShowIou contract on which the painter is a stakeholder. This now requires an additional contract type, compared to the original paint offer example. An alternative approach to enable this workflow, without increasing the number of contracts required, is to replace the original Iou contract by one on which the painter is a contract observer. This would require extending the contract model with a (consuming) exercise action on the Iou that creates a new Iou, with observers of Alice’s choice. In addition to the different number of commits, the two approaches differ in one more aspect. Unlike stakeholders, parties who see contracts only through divulgence have no guarantees about the state of the contracts in question. For example, consider what happens if we extend our (original) paint offer example such that the painter immediately settles the IOU.

A series of time sequences in which the state of the Iou contract is unknown to parties who view it through divulgence.

While Alice sees the creation of the Iou Bank P contract, she does not see the settlement action. Thus, she does not know whether the contract is still active at any point after its creation. Similarly, in the previous example with the counteroffer, Alice could spend the IOU that she showed to the painter by the time the painter attempts to accept her counteroffer. In this case, the painter’s transaction could not be added to the ledger, as it would result in a double spend and violate validity. But the painter has no way to predict whether his acceptance can be added to the ledger or not.