Re: Yet another intractible pathing problem
Posted: Fri Aug 22, 2008 11:50 pm
Sometimes I wonder whether I am being a little too harsh on the people from RSDL -after all, they seem like pleasant people, they take a lot of their time to post on this forum, they do take into account community feedback when releasing upgrades, they do release good free addon content from time to time, and the payware addons seem to be good, too, although I have not bought any, and at least one of them has good taste in television sitcoms. However, when the commercial product that it is their raison d'etre to maintain is cripplingly defective, those crippling defects dominate and obliterate everything else. Any signs that it is not being taken as seriously as it should be are matters for very great concern indeed.
Imagine, in the early days of mass produced automobiles, that you bought a car, beautifully crafted, comfortable to drive, with a powerful engine - much better in those respects than anything that had gone before, or, indeed, is available now. The car had been advertised much in advance as being capable of high speeds. When you drive it away, you find that it will only work in first gear, and the clutch often sticks. You take it back to the garage only to be told that the original manufacturers have nothing to do with it, and that all maintenance has been subcontracted to a tiny little garage. On arriving at the tiny little garage, the two friendly mechanics explain that the manufacturer didn't have time to perfect the transmission, but that the car can still do everything that you want it to do, and refer you to a very small owner's club, half of whose members are complaining bitterly about the transmission, and half of whose members only want the car to drive slowly up and down the road outside their house, and don't terribly mind that it doesn't change out of first gear. When you ask how you reverse it, and what to do if you need to get out of a tight parking spot, you are told that you need to be aware of the car's limitations and that it is best only to drive it where there is plenty of room to turn around without reversing. When you ask how you are supposed to drive it at high speeds over long distances, you are told that you are really expecting a lot from the car, and that one shouldn't look back on what could have been, and that we are where we are, and by the way, would you like to buy one of our special upgraded headlamps or a different colour seat cover?
You point out that the very low sales of the car might well be attributable to the poor transmission, but are referred to the owners' club, and told that many of them are happy with it. When you suggest that there would be a huge market for a car like that with working transmission amongst the majority of people who want it to do more than drive up and down at low speed, you get a blank look and no reply. You also get no answer when you ask for a comprehensive guide to the functions of the gearbox and the various inexplicable levers that seem to surround it, but instead are referred to the owner's manual that only refers to half the levers and explains their function with varying degrees of depth.
You take the gearbox apart, and find that, as it is designed, it will never be capable of working in anything other than one gear, even though it was evidently intended to have four forward gears and a reverse. You point this out to the tiny garage, who don't reply. You notice that they don't seem to have anybody there who knows anything about how to design new systems or set up the tooling to produce entirely new parts. The car has been serviced a few times, but always in ways that involved a little work from a fitter or electrician: no newly designed component parts have ever been added. You ask the little garage whether they can ever fix the gearbox issues, and one of the friendly mechanics replies that they know that there are a few limitations with the transmission, but that they're working on doing something about it for the next service. You remember some time ago the same mechanic mentioning that the transmission would be at the top of the list of priorities for the previous service, but all that they did was to tighten the clutch cable and put a higher grade oil in the gearbox.
You are still taking the 'bus to work every day despite having had the car for nearly a year. You have only driven it a few times, and, apart from those occasions when you just drove it up and down the road outside your house to test drive it, have had to push the car home or have it towed. In those circumstances, one might be forgiven for being more than a little harsh on the two mechanics, however friendly that they might be.
I will deal with some of the specific issues raised in this thread by subheading.
Complexity of the scenario
Some have suggested that the scenario that I was trying to create was a complicated one: let me emphasise that, at the stage that I have reached so far, it is extremely simple: there are, in effect, only two diagrams (with some minor variations, such as whether they stop at Slough or which platform at Paddington that they take): (1) London-Reading-Baisingstoke portal; and (2) London-Reading-Didcot-Swindon portal. In reality, that route would have well over twenty diagrams. It seems that it cannot cope with the fact that the Baisingstoke trains cross the path of the Didcot trains at Reading West junction (a fundamentally basic railway operation) because of the enormous number of permutations that the pathing system, given its absurd design, has to calculate for even such a simple manuvre.
Some people have had success with scenarios with a very large number of AI trains. I have not seen those scenarios, as they have not, to my knowledge, been made available to the public, but I should note that, in my observation, the overall number of trains is less important than the possibility of conflict. I had set up on my route a goodly number of trains on the London-Swindon diagram with no trouble: the only problems came when I started adding even a few trains to the Baisingstoke/Westbury diagram.
"Working within Rail Simulator's limitations" means hardly working at all - given that setting up only two simple diagrams on the Oxford to Paddington route (not even making any use of the relief lines at all) causes the system to fail completely, the limitations of Rail Simulator are equivalent to those of the car that only works in first gear described above. References to finding one's way around a brick wall are of no use if there is no way around the wall because of its vast size: that is the case in many respects in Rail Simulator. How would one overcome the "brick wall" of this particular problem, when, having intended to create a scenario with dozens of diagrams, the pathing engine fails completely after creating only two?
Portals
From my understanding, trains are not supposed to enter portals and then re-emerge from them again: entry into a portal is supposed to make a train vanish for ever. If one wants a train to enter and then another train of the same sort to emerge later, one should use two trains. It ought to be possible to manage the trains stored in portals more easily, including allowing specific trains that have entered to re-emerge, but that is not possible. This, however, at least, is not a crippling flaw in the way that the instability that I describe in the original post is.
If we recall, the portal stability issues were supposed to have been fixed in Upgrade Mk 1. I am not convinced that portals are causing my problems (which I still think are caused by the failure to handle conflicts at Reading West Junction without entering an infinite loop because of the impossible number of permutations created by the absurd brute force approach of the pathing engine), but if people put a portal as a non-final destination, and then have another instruction after a portal, that might well be causing the trouble, since I do not think that portals were designed to work that way (although I might be wrong: I have never tried it. Perhaps that is what the values attached to the portal in the route edit mode are about).
I do note, however, that I put all the trains into the Swindon portal and never had any difficulties with that: it was only when I was setting up trains from the Basingstoke portal, crossing the paths of trains from the Swindon portal, that I had difficulties. If anyone can run a test that clearly establishes that a particular problem is caused specifically by a portal, however, please let me know, and I will mark the issue as not fixed on my list of critical issues. That would be significant, since it would be an example of an issue that RSDL have claimed to fix, but that has not, in fact, been fixed, which would be reason for even greater scepticism that any of the serious issues will ever be fixed.
One final thing to note about portals is this: when a train is placed on the track just outside a portal (which it has to be in order for its timetable to be set up), it is blocking the track out of the portal, so all of the trains currently in the portal will be unable to find a path if they are scheduled to leave later than the train on which one is currently working. Always, therefore, add the train to the portal before checking the drivers tab to see whether any routes have failed. Similarly, when setting up a train's timetable, finish it before checking to see whether the train conflicts with anything else, or else one will be testing for conflicts based on the train stopping part way through its timetable and never moving again, which will likely block other trains.
Forum discussions and individual help
Whilst it is certainly quite a privilege to be able to receive personal help from somebody who has worked on the development of the software, it is far more useful generally if the advice can be made public so that everybody with a similar problem can see how to solve it, without asking the same question over and over (and without having to wait for a reply). Indeed, anything of this nature with which people are helped privately ought really be included in the official documents.
I do not understand why any official representative of a software producer should want to prioritise helping people one at a time rather than providing information that would help lots of people at the same time. There is no conflict between the two: the same conversation as takes place in the private e-mails could equally take place in the public forum. It is a false dichotomy to distinguish between, on the one hand, helping people in private e-mails, and, on the other, explaining on public forums why decisions have already been taken: there is nothing to stop one from helping individuals in a public forum where others with the same or similar problems can share in the help, or publishing documents that contain all the useful information given out to people one by one in e-mails so as to reduce the need for people asking for help in the first place?
But, can I ask, if I had written to you and/or Adam and asked about this pathing problem, would you or he have had a solution to it?
In any event, when I refer to specifics not being discussed, I am referring, not to requests for help, but to threads, such as this one, pointing out the defects of the software. The public announcements of what will be done, and the public statements about how to interact with the system as it is now, and what its limitations are, are, without exception, entirely devoid of detail.Why on earth is there not a document somewhere setting out exactly how the pathing engine works, what its problems and limitations are, and how to set up scenarios without things going wrong? Such a document would undoubtedly save a huge number of people an enormous amount of time and wasted effort creating things that the software is quite incapable of dealing with. What kind of marketing strategy is it to let people become so frustrated with the product, by not giving any information about its manifest limitations, that they are likely to cease using it, never buy any addons or expansion packs, tell all of their friends never to buy it, and tell them to tell all of their friends never to buy it (or anything with which the developer in question has ever been involved)? I am concerned that the reason might be that any such document would have to admit that the limitations were so severe as would dissuade people from buying the game in the first place. It would certainly be an obvious contradiction of the dishonest claim made on the website about "completely accurate local signalling systems with all the bells and whistles".
Acknowledgement of problems
I note that Derek above notes that "the dispatcher has a few issues finding paths", and then goes on to suggest that it is not always that dispatcher's fault that it does not find a path. Two points: firstly, as I have written elsewhere, the fact that a signal set up incorrectly or properties in the track, etc., causes the pathing engine to fail in the way that it does is itself an indication that something is seriously wrong. No matter how badly set up the route or scenario, the software should never hang or crash or behave unpredictably. There should be no infinite loops for any possible scenario that the user could create. It should not be possible to put that pathing engine under so much strain by any sane use that the reported status in the drivers tab gets stuck on "pending". One of the main problems with the pathing engine is precisely that it fails utterly to deal with difficult situations or handle problems in an intelligent way. To state that these failures are not the pathing engine's fault because something else has given it a difficult set of parameters to deal with entirely misses the point.
Secondly, there is a great irony in replying to a post explaining concerns at not taking seriously the fact that the pathing engine is cripplingly defective and fundamentally broken by stating, "we are not denying that the dispatcher has a few issues finding paths that we need to deal with" (emphasis mine). The whole point was not that there has been no acceptance that there are any problems with pathing, but that there has been no acceptance that the problems with pathing are both serious and fundamental. This leads to the concern, which has already been expressed in relation to Upgrade Mk 2 (in which it was originally promised that signalling would be "near the top of the list of priorities"; it was only much later that it was stated that the dispatcher would not be fixed in Mk2), that any fixes to the pathing engine will be too superficial to have any noticable impact on the real problems. To that is added the concern (as expressed by AndyUK in relation to physics) that there are simply not enough resources at RSDL (whether in terms of numbers or expertise or both) to undertake the sort of ground up rewriting of the code necessary to have more than a negligible impact on these issues.
Imagine, in the early days of mass produced automobiles, that you bought a car, beautifully crafted, comfortable to drive, with a powerful engine - much better in those respects than anything that had gone before, or, indeed, is available now. The car had been advertised much in advance as being capable of high speeds. When you drive it away, you find that it will only work in first gear, and the clutch often sticks. You take it back to the garage only to be told that the original manufacturers have nothing to do with it, and that all maintenance has been subcontracted to a tiny little garage. On arriving at the tiny little garage, the two friendly mechanics explain that the manufacturer didn't have time to perfect the transmission, but that the car can still do everything that you want it to do, and refer you to a very small owner's club, half of whose members are complaining bitterly about the transmission, and half of whose members only want the car to drive slowly up and down the road outside their house, and don't terribly mind that it doesn't change out of first gear. When you ask how you reverse it, and what to do if you need to get out of a tight parking spot, you are told that you need to be aware of the car's limitations and that it is best only to drive it where there is plenty of room to turn around without reversing. When you ask how you are supposed to drive it at high speeds over long distances, you are told that you are really expecting a lot from the car, and that one shouldn't look back on what could have been, and that we are where we are, and by the way, would you like to buy one of our special upgraded headlamps or a different colour seat cover?
You point out that the very low sales of the car might well be attributable to the poor transmission, but are referred to the owners' club, and told that many of them are happy with it. When you suggest that there would be a huge market for a car like that with working transmission amongst the majority of people who want it to do more than drive up and down at low speed, you get a blank look and no reply. You also get no answer when you ask for a comprehensive guide to the functions of the gearbox and the various inexplicable levers that seem to surround it, but instead are referred to the owner's manual that only refers to half the levers and explains their function with varying degrees of depth.
You take the gearbox apart, and find that, as it is designed, it will never be capable of working in anything other than one gear, even though it was evidently intended to have four forward gears and a reverse. You point this out to the tiny garage, who don't reply. You notice that they don't seem to have anybody there who knows anything about how to design new systems or set up the tooling to produce entirely new parts. The car has been serviced a few times, but always in ways that involved a little work from a fitter or electrician: no newly designed component parts have ever been added. You ask the little garage whether they can ever fix the gearbox issues, and one of the friendly mechanics replies that they know that there are a few limitations with the transmission, but that they're working on doing something about it for the next service. You remember some time ago the same mechanic mentioning that the transmission would be at the top of the list of priorities for the previous service, but all that they did was to tighten the clutch cable and put a higher grade oil in the gearbox.
You are still taking the 'bus to work every day despite having had the car for nearly a year. You have only driven it a few times, and, apart from those occasions when you just drove it up and down the road outside your house to test drive it, have had to push the car home or have it towed. In those circumstances, one might be forgiven for being more than a little harsh on the two mechanics, however friendly that they might be.
I will deal with some of the specific issues raised in this thread by subheading.
Complexity of the scenario
Some have suggested that the scenario that I was trying to create was a complicated one: let me emphasise that, at the stage that I have reached so far, it is extremely simple: there are, in effect, only two diagrams (with some minor variations, such as whether they stop at Slough or which platform at Paddington that they take): (1) London-Reading-Baisingstoke portal; and (2) London-Reading-Didcot-Swindon portal. In reality, that route would have well over twenty diagrams. It seems that it cannot cope with the fact that the Baisingstoke trains cross the path of the Didcot trains at Reading West junction (a fundamentally basic railway operation) because of the enormous number of permutations that the pathing system, given its absurd design, has to calculate for even such a simple manuvre.
Some people have had success with scenarios with a very large number of AI trains. I have not seen those scenarios, as they have not, to my knowledge, been made available to the public, but I should note that, in my observation, the overall number of trains is less important than the possibility of conflict. I had set up on my route a goodly number of trains on the London-Swindon diagram with no trouble: the only problems came when I started adding even a few trains to the Baisingstoke/Westbury diagram.
"Working within Rail Simulator's limitations" means hardly working at all - given that setting up only two simple diagrams on the Oxford to Paddington route (not even making any use of the relief lines at all) causes the system to fail completely, the limitations of Rail Simulator are equivalent to those of the car that only works in first gear described above. References to finding one's way around a brick wall are of no use if there is no way around the wall because of its vast size: that is the case in many respects in Rail Simulator. How would one overcome the "brick wall" of this particular problem, when, having intended to create a scenario with dozens of diagrams, the pathing engine fails completely after creating only two?
Portals
From my understanding, trains are not supposed to enter portals and then re-emerge from them again: entry into a portal is supposed to make a train vanish for ever. If one wants a train to enter and then another train of the same sort to emerge later, one should use two trains. It ought to be possible to manage the trains stored in portals more easily, including allowing specific trains that have entered to re-emerge, but that is not possible. This, however, at least, is not a crippling flaw in the way that the instability that I describe in the original post is.
If we recall, the portal stability issues were supposed to have been fixed in Upgrade Mk 1. I am not convinced that portals are causing my problems (which I still think are caused by the failure to handle conflicts at Reading West Junction without entering an infinite loop because of the impossible number of permutations created by the absurd brute force approach of the pathing engine), but if people put a portal as a non-final destination, and then have another instruction after a portal, that might well be causing the trouble, since I do not think that portals were designed to work that way (although I might be wrong: I have never tried it. Perhaps that is what the values attached to the portal in the route edit mode are about).
I do note, however, that I put all the trains into the Swindon portal and never had any difficulties with that: it was only when I was setting up trains from the Basingstoke portal, crossing the paths of trains from the Swindon portal, that I had difficulties. If anyone can run a test that clearly establishes that a particular problem is caused specifically by a portal, however, please let me know, and I will mark the issue as not fixed on my list of critical issues. That would be significant, since it would be an example of an issue that RSDL have claimed to fix, but that has not, in fact, been fixed, which would be reason for even greater scepticism that any of the serious issues will ever be fixed.
One final thing to note about portals is this: when a train is placed on the track just outside a portal (which it has to be in order for its timetable to be set up), it is blocking the track out of the portal, so all of the trains currently in the portal will be unable to find a path if they are scheduled to leave later than the train on which one is currently working. Always, therefore, add the train to the portal before checking the drivers tab to see whether any routes have failed. Similarly, when setting up a train's timetable, finish it before checking to see whether the train conflicts with anything else, or else one will be testing for conflicts based on the train stopping part way through its timetable and never moving again, which will likely block other trains.
Forum discussions and individual help
Whilst it is certainly quite a privilege to be able to receive personal help from somebody who has worked on the development of the software, it is far more useful generally if the advice can be made public so that everybody with a similar problem can see how to solve it, without asking the same question over and over (and without having to wait for a reply). Indeed, anything of this nature with which people are helped privately ought really be included in the official documents.
I do not understand why any official representative of a software producer should want to prioritise helping people one at a time rather than providing information that would help lots of people at the same time. There is no conflict between the two: the same conversation as takes place in the private e-mails could equally take place in the public forum. It is a false dichotomy to distinguish between, on the one hand, helping people in private e-mails, and, on the other, explaining on public forums why decisions have already been taken: there is nothing to stop one from helping individuals in a public forum where others with the same or similar problems can share in the help, or publishing documents that contain all the useful information given out to people one by one in e-mails so as to reduce the need for people asking for help in the first place?
But, can I ask, if I had written to you and/or Adam and asked about this pathing problem, would you or he have had a solution to it?
In any event, when I refer to specifics not being discussed, I am referring, not to requests for help, but to threads, such as this one, pointing out the defects of the software. The public announcements of what will be done, and the public statements about how to interact with the system as it is now, and what its limitations are, are, without exception, entirely devoid of detail.Why on earth is there not a document somewhere setting out exactly how the pathing engine works, what its problems and limitations are, and how to set up scenarios without things going wrong? Such a document would undoubtedly save a huge number of people an enormous amount of time and wasted effort creating things that the software is quite incapable of dealing with. What kind of marketing strategy is it to let people become so frustrated with the product, by not giving any information about its manifest limitations, that they are likely to cease using it, never buy any addons or expansion packs, tell all of their friends never to buy it, and tell them to tell all of their friends never to buy it (or anything with which the developer in question has ever been involved)? I am concerned that the reason might be that any such document would have to admit that the limitations were so severe as would dissuade people from buying the game in the first place. It would certainly be an obvious contradiction of the dishonest claim made on the website about "completely accurate local signalling systems with all the bells and whistles".
Acknowledgement of problems
I note that Derek above notes that "the dispatcher has a few issues finding paths", and then goes on to suggest that it is not always that dispatcher's fault that it does not find a path. Two points: firstly, as I have written elsewhere, the fact that a signal set up incorrectly or properties in the track, etc., causes the pathing engine to fail in the way that it does is itself an indication that something is seriously wrong. No matter how badly set up the route or scenario, the software should never hang or crash or behave unpredictably. There should be no infinite loops for any possible scenario that the user could create. It should not be possible to put that pathing engine under so much strain by any sane use that the reported status in the drivers tab gets stuck on "pending". One of the main problems with the pathing engine is precisely that it fails utterly to deal with difficult situations or handle problems in an intelligent way. To state that these failures are not the pathing engine's fault because something else has given it a difficult set of parameters to deal with entirely misses the point.
Secondly, there is a great irony in replying to a post explaining concerns at not taking seriously the fact that the pathing engine is cripplingly defective and fundamentally broken by stating, "we are not denying that the dispatcher has a few issues finding paths that we need to deal with" (emphasis mine). The whole point was not that there has been no acceptance that there are any problems with pathing, but that there has been no acceptance that the problems with pathing are both serious and fundamental. This leads to the concern, which has already been expressed in relation to Upgrade Mk 2 (in which it was originally promised that signalling would be "near the top of the list of priorities"; it was only much later that it was stated that the dispatcher would not be fixed in Mk2), that any fixes to the pathing engine will be too superficial to have any noticable impact on the real problems. To that is added the concern (as expressed by AndyUK in relation to physics) that there are simply not enough resources at RSDL (whether in terms of numbers or expertise or both) to undertake the sort of ground up rewriting of the code necessary to have more than a negligible impact on these issues.
Although i take and accept your points of the problems that RS suffers with and they are pretty frustrating indeed, there is still enjoyment to be had if one is not quite as technical as one may be inclined, and being one whom is baffled by the technical 
, (even in simplicity) doesn't mind the simple pleasure of A to B scenario creation (see my Vintage trains scenario 
) however i never use portals to spawn Ai, only to get rid of them, i always use spawn yards or sidings to start my services from which for me seems to work very well. 
as you say ignoring red signals even though the driver list shows all is well! 
The claims for KRS were pretty lofty at the time of release which has not been met by the reality, particularly where the signaling is concerned! (you can see on other threads about this on these forums) If somebody sells you something that doesn't do what it claims then you are going to be a tad upset, no? And as an example is it right when passing a green signal to be confronted with a red signal 0.3 miles ahead and doing 90 miles per hour after Mk2 upgrade?