Note: the discussion in this thread concerns a suggested feature set that there is unlikely to be time to implement for many years or even decades unless somebody else be willing to invest very considerable time in working on this. The current priority, aside from fixing any critical bugs, is to complete the balance critical features and then to balance the pakset properly.
The pandemic of 2020 has accelerated trends that have been ongoing for some time. It is now possible to see a future in which the rate of commuting per unit of population will be permanently much lower for any given rate of unemployment than it has ever been in history owing to the rise of home working. Likewise, there will be many people for whom physical distance to work becomes irrelevant and many jobs that will be able to be fulfilled by people in any location providing that both employer and employee have a good internet connexion. Likewise, in person visits to shops are rapidly being replaced by online shopping, and for many years e-mail and other electronic communication methods have largely replaced traditional paper based post for written communication.
These changes have a significant impact on the economic model for transport simulated in Simutrans-Extended. The current model assumes a fixed commuting and visiting rate, and that passengers will travel if they can. The planned model for town growth assumes that people need to be able to commute to work within a certain time in order to want to live in a particular place: an assumption that has held true so far, but which may well not be true for a significant number of people in the near future.
To model the transportation of the mid 21st century as well as we plan to model the transportation of the 18th-20th centuries, therefore, we need to be able to model the fact that some functions of passenger and mail travel can be fulfilled by alternative means (e.g. e-mail, telephone, video conferencing or even older methods such as the telegram or telex).
It is too early at this stage to design the system in detail, but some basic ideas emerge at this preliminary stage. First of all, to do this properly, it is desirable to model electronic communications networks directly rather than to abstract them. This can in principle be done in approximately the same way as with the electricity network, and some code may even be shared. Modelling electronic communication networks is no further from the core function of Simutrans-Extended of modelling transport than is modelling electricity: indeed, in many ways, communications are closer in function to transport than electricity. Communications networks modelled can include telegraph, telephone and internet (and the latter in varying bandwidths so as to distinguish between a connexion good enough for e-mail and a connexion good enough for video conferencing). Similarly to electricity, it makes sense for the direct modelling to be limited to the trunk networks and to abstract the networks internal to towns: for example, placing a single telephone exchange in a town might automatically allow the whole town to communicate with other people in the town by telephone according to a per wealth level and per date table of telephone adoption rates; but calls between towns would only be possible if the towns were linked by telephone wires actually laid by the player. Consideration would have to be given to what happens if two different players build telephone exchanges in the town. Some complex load balancing code may be required.
Secondly, consideration will have to be given to how to simulate network capacity dynamically. This is likely to be very challenging technically and may require substantial mathematical wizardry to make work efficiently (especially since the capacity of a telephone network works differently to the capacity of the internet). We currently do not simulate the capacity of the electricity network itself: only generation capacity. We assume that the electrical wires have an unlimited ability to carry current. However, we cannot get away with this for communication networks in respect of which capacity is a critical element of the economics.
Thirdly, consideration will need to be given to the algorithm for determining when a possible passenger or mail trip can be fulfilled by alternative means. What proportion of the time in the past did people physically visit others when a telephone call would have been considered by them to be sufficient had the technology been available? What proportion of letters sent in 1990 would be sent by e-mail by anyone who had the ability to do so in 2000? We may need to break down the passenger trips into further subtypes beyond visiting and commuting so as to simulate which kinds of trips can be satisfied by alternative fulfilment given the current passenger's access to technology and the current free capacity on the network; or alternatively, it may suffice to randomise this. The former may lead to a great increase in complexity, however, and the latter to distortions, so very careful thought will have to be given to this.
Fourthly, we will have to consider the effect of alternative fulfilment on Marchetti's Constant. Will people who can telecommute make more visiting trips as the model predicts? If so, we will need to consider how to model this without breaking the existing modelling of Marchetti's Constant, which is not easy to adapt dynamically (although some basic dynamic adaptation has been added recently as discussed elsewhere).
Fifthly, we will need to consider the interaction between this and consumer industries; we may need to invent a whole new home delivery mechanic for consumer industries (and that might link in with the planned private vans feature - another far distant feature long beyond current balancing goals), and that would be a topic deserving of a thread in its own right so complex are the ramifications of that. Of course, this would then allow more traditional forms of home delivery popular in the past, such as milk floats and local bicycle deliveries of groceries. Whether to do this by catchment area or actual pathfinding vehicles will need some thought. A similar model to private cars might be used for such vehicle trips (indeed, much of the same code and even data could be used), but perhaps with different vehicle graphics and statistics, depicting slow milk floats and faster delivery vans, which might share movement code with private cars rather than player vehicles (save that they might not despawn on reaching their destination, but might instead make an immediate return trip).
Bearing in mind that implementation of this feature might take significantly longer than the actual economy takes to move to long-term home working for all or most office workers, I should be grateful for any feedback on this.
