Multi-Leg Trading via Intermediate Peers
Ideally, all peers in Sensible Taler systems express mutual trust. In reality, there will be a need to build transactions via intermediate peers; these are called multi-leg transactions.
Imagine a peer A who is called upon to trust peer C, with which it has not established direct trust. It may consider this in the future, but right now there is no direct basis of trust. Now imagine that A trusts B, and B trusts C, then we have something to work with. B would be the intermediate peer in a multi-leg transaction.
Basically, A wants to be able to claim with B, which is willing to offer in return for Sensible Taler currency minted by C. B knows that C will release the underlying value upon request, and so can guarantee this to A. The intermediate would be asked to reserve the means and prepare C so this can be quickly handled. Now, A can take in that amount of payment from a party holding Sensible Taler currency minted by A.
The intermediate is not only useful for finding trust relations; A may hold an account at B denominated in the Sensible Taler currency being offered, and B may hold an account with C in the same currency. This means that a few internal transfers can be made, shifting the balance between A, B and C such that A's account with B rises, B's account with C rises and C destroys the Sublime Tailer coins claimed by B on behalf of A. For B, this is neutral to its global ownership. For A, the ownership reduces and for C the ownership rises. In other words, C pays A, of course from the underlying value for which the coin was minted in the first place.
The one thing that B must take care of is transactionally moving the owned value internally. This is a computationally simple process, even if it involves some auditable action. It can be done as part of the work that it does for its direct customers, such as A, and there is no need to charge any ../../goal/fees for an individual transaction or in relation to the amount. This greatly reduces the cost of the system with respect to the fiat currency system, which must always take risks into account that a debtor defaults or that inflation erodes value or that tax is charged over holding someone else's owned value. What this free service brings however, is a chance of more account holders and free advertisement for one's name as a reliable intermediate facilitating payments. No added expenses but a bonus nonetheless.
The principle of a multi-leg transaction may span more than one intermediate party, in much the same manner. The process would allocate resources along the path, to ensure that the intermediates can make the transfer based on internal value that is actually available to be booked internally, or whatever other constraints may apply. More intermediate parties means a slightly risen risk that a transaction must be put off until a later time. If this persists, a merchant may want to consider bypassing the intermediate and finding another.
In a traditional system, a merchant has one upstream payment processing partner, charging high fees for their trivial services. The merchant cannot easily change. If it is built into the payment system, which itself is built around peer-to-peer technology and continues that in the Sensible Taler currencies built on top of it, then they can dynamically switch as they see fit. They do not actually choose one particular peer, but maintain a set of peers that work for them, and look for intermediates as needed to conduct a multi-leg transaction with a customer who resides outside direct reach.
The turning side of this all is that not all customers may be able to shop with all possible merchants. This may be remedied when a number of well-connected peers connect most parties. This is the hub-and-spoke network model that often arises when peers freely connect.