it's on Github. You can check the code there.
But I had problems understanding the "old" convoy-code too.. :-)

Thanks for the announcement.

Though of course I tried to find a sufficient set of test cases there can always be some bugs left, especially as my set of different devices and browsers is very limited. So if you observe any problems like convoy orders that you cannot set or invalid convoy orders that you can set or an irresponsive website / site crashes do not hesitate to report them. So I could try to fix those bugs as soon as possible.

While working on the convoy code, Enriador and I prepared some more variants for the interactive map, as well. Many thanks to Enriador here who is much more effective than me in preparing the maps and gives me hope, that we will finally complete the list of variants with interactive map support. And who apparently finds time to do this stuff between producing one variant after another. Very impressive! :-)

So for those like YCHTT who might be interested in the algorithm: The main problem with the convoy code weren't the simple convoy path searches from A to B but more complex ones A to B via C. Those searches / checks are needed to see if a fleet at C can actually convoy an army from A to B. The original algorithm for this as well as for the normal A to B problem was very basic and literally searched for *every* simple path from A to B until it find one that passes through C, as well.
This basic approach is sufficient for small maps like the classic one. But issues are even known for webdip's World variant. On bigger variants like the WWIV ones the path searches and checks were just deactivated so you could enter invalid convoys here (though that wouldn't pass the adjudicator). But fortunately there was much room for optimization.
I followed a first approach a few weeks ago that reduces the number of paths that were searched. But that was not enough for the extreme test case of a WWIV sealanes map with fleets on every sea province. So I put more thoughts into the problem and am confident to claim that I found an efficient way to search a small set of paths from A to C (or B to C depending on what is further away) that can be labeled as relevant routes. This in the end leads to algorithm that is producing correct results even on the extreme sealanes test case in just a few 10 ms.
However, I do not want to go into more detail here but - as Oli did - I will point to the Github repository. I tried to add a sufficient documentation so somebody else can actually work on my code. So I hope the interested ones will be able to understand the algorithm in detail. You can find all changes here: https://github.com/Sleepcap/vDiplomacy/blob/master/javascript/board/model_vDipV2.js

I'm not a Perl or PHP guy but it is similar enough to C++ I should be able to follow it. Thanks for the effort!