Looking for feedback on my new split ticket finding program

[Splitix]

Over the past few years I've been writing a little program for finding split train tickets. I think it's finally good enough to release to the world.

Is there any one here who wouldn't mind taking it for a spin and letting me know what they think?

The program is focused around my needs and use cases - so I'd be particularly interested to hear how generally applicable and useful anyone thinks it is!

So without further ado - http://www.splitix.co.uk

 

[trainfanatic]

I think there's something wrong with your programme......

 

[Nym]

Appears to work reasonably nicely for what it's intended for. However, I think a few words of re-assurance may be in order for those suspicious of new software...

 

[4SRKT]

I think you might find considerable hostility on here to the notion of something that makes split ticketing easy and widely used/understood. If everyone uses it then the obvious reaction of the TOCs will be to even out fares so that it no longer makes sense to split.

 

[Nym]

I think you might find considerable hostility on here to the notion of something that makes split ticketing easy and widely used/understood. If everyone uses it then the obvious reaction of the TOCs will be to even out fares so that it no longer makes sense to split.

Quite, and it will be noticed, either for being indexed by google, or this forum, or by the immense amount of traffic is causes for the NR website.

 

[headshot119]

I think there's something wrong with your programme......

Which is?

 

[bb21]

In the best splits section, the Southampton-Waterloo example is incorrect. The comparison seems to be railcard-discounted fares against full fares.

 

[Nym]

Well, I can see two or three major problems with the actual software, but go on...
--- old post above --- --- new post below ---

In the best splits section, the Southampton-Waterloo example is incorrect. The comparison seems to be railcard-discounted fares against full fares.

It is indeed, but there's a few more fundamental problems, but I want to see if mr critical from the top of the page can come up with it.

 

[bb21]

Well, I can see two or three major problems with the actual software, but go on...
--- old post above --- --- new post below ---


It is indeed, but there's a few more fundamental problems, but I want to see if mr critical from the top of the page can come up with it.

Same with Doncaster-Nottingham, and I have only checked two examples so far, so both flawed. Some further work is needed I reckon.

Some of the best savings don't actually come from splits, especially for short- to medium-distance journeys. They tend to take advantage of RG anomalies, and that is not something that can be computerised.

 

[headshot119]

Major flaw in the software, asking for Wrexham General - Chester seems to break it.

It also doesn't factor in over distance.

And some of the other splits it finds aren't the cheapest. A prime example asking it for Wrexham General - Newport it says split at Ruabon, that certainly isn't the cheapest split.

 

[Nym]

So the more fundamental issues of it being like a DoS Attack on nationalrail.co.uk to get it to work, unbelievably simplistic algorithms making this worse, reporting all of these back to their own site to log every search that is conducted (I'm now running it visualised behind a firewall), not accounting for over distance, not supporting return fares, and not telling you what type of fares to get for each ticket aren't noticeable?

It's a, I won't say simple because it isn't, but it's an implementation of nodal analysis with dijkstra's algorithm that takes into account stations called for that specific route given by NR for any given search, so it's far from perfect, but is a handy tool. (if you block the reporting)

 

[nedchester]

Seems a good program.

 

[Splitix]

Wow, plenty of good feedback so far. Let's see if I can tease out what the problems are:

@trainfanatic - Can you elaborate?
@Nym - I'm aware that the website is a bit "sparse" at the moment. I will add some reassurance in due course. Can you elaborate on the other major problems? I would like to fix them...
@4SRKT - Is split ticketing not allready quite widely known? Sites such as moneysavingexpert.com have featured it for quite a while already and they have >million people on their mailing list.
@bb21 - I will look into this... (http://www.splitix.co.uk/stations/Southampton Central/London Waterloo). Are the problems you're seeing just on the results on the website, or from using the downloaded program?
@headshot119 - I will look in to the Wrexham General->Chester problem in a minute. What do you mean by "over distance"? And for http://www.splitix.co.uk/stations/Wrexham General/Newport (South Wales) - can you give me an example of a cheaper split?

 

[headshot119]

Wow, plenty of good feedback so far. Let's see if I can tease out what the problems are:


@headshot119 - I will look in to the Wrexham General->Chester problem in a minute. What do you mean by "over distance"? And for http://www.splitix.co.uk/stations/Wrexham General/Newport (South Wales) - can you give me an example of a cheaper split?

I'll retract my earlier comment about Wrexham General - Newport, the split I used to use has climbed in price, so the Ruabon split is now cheaper.

Regarding WRX - CTR it seems to just repeat the last journey I asked for, before the WRX - CTR.

Also it doesn't factor in the use of Rover and Ranger tickets.

 

[Paul Kelly]

It is a Java GUI app that screen-scrapes the National Rail Enquries website to get timetable and fare information, then does some local analysis on your own computer looking for basic splits along the calling points of the chosen train, and posts the results back to its own website. I think it's pretty good for a first attempt. But you're probably not allowed to use the NRE site like that...

 

[Nym]

For a bit more elaboration, I logged over 8000 packets associated with nationalrail.co.uk when searching for a split...

 

[Splitix]

It is a Java GUI app that screen-scrapes the National Rail Enquries website to get timetable and fare information, then does some local analysis on your own computer looking for basic splits along the calling points of the chosen train, and posts the results back to its own website. I think it's pretty good for a first attempt. But you're probably not allowed to use the NRE site like that...

Yep - that's an accurate summary

 

[Nym]

It is a Java GUI app that screen-scrapes the National Rail Enquries website to get timetable and fare information, then does some local analysis on your own computer looking for basic splits along the calling points of the chosen train, and posts the results back to its own website. I think it's pretty good for a first attempt. But you're probably not allowed to use the NRE site like that...

You're not no, and as above, I don't think ATOC would be happy about that many enquires in one go. And btw, it's YOUR IP address making the enquires, so it's YOU that DBS/EWS (The people who run NR's servers) will be annoyed with...

PS: I know at least two people on here will have already de-compiled it

 

[4SRKT]

@4SRKT - Is split ticketing not allready quite widely known? Sites such as moneysavingexpert.com have featured it for quite a while already and they have >million people on their mailing list.

Yes it is, but splitting is actually quite complex in practice and many people I know don't do it unless I explicitly explain what to ask for. Even excessing goes over many people's heads: people just preferring to pay the extra single fare rather than try and get their head around the concept of excess fares. Moneysavingexpert doesn't provide to my knowledge an actual algorithm for working it out, merely saying it's possible and giving examples. This moves splitting onto the next level.

The reaction of a TOC such as XC with very expensive long distance walk on fares could well be this. Say (for example) an anytime Leeds > Birmingham ticket is about £90, and people more than halve it by routinely splitting at Sheffield and Derby, they will have to make up the revenue from somewhere. The losers will be the people actually travelling from Sheffield to Derby, or Derby to Birmingham, whose fares will have to rise (at the same time as the Leeds > Birmingham fare falling) to fill the gap. Not just a few enthusiasts disbenefitting in that case.

I'm impressed on one level by what you're trying to do, but worry about it on another. I had thought about trying to set up such a venture myself once, commercially, but realised that it would be a short-lived venture with no raison d'etre once the TOCs acted to defend themselves against it.

 

[Splitix]

So the more fundamental issues of it being like a DoS Attack on nationalrail.co.uk to get it to work, unbelievably simplistic algorithms making this worse, reporting all of these back to their own site to log every search that is conducted (I'm now running it visualised behind a firewall), not accounting for over distance, not supporting return fares, and not telling you what type of fares to get for each ticket aren't noticeable?

It's a, I won't say simple because it isn't, but it's an implementation of nodal analysis with dijkstra's algorithm that takes into account stations called for that specific route given by NR for any given search, so it's far from perfect, but is a handy tool. (if you block the reporting)

Thanks for the detailed look I've no immediate plans to add support for return fares, but may do in the future. I can and will add details of what type of fare to purchase for each ticket - I can't believe I missed that to begin with!

I thought over distance wasn't allowed?
--- old post above --- --- new post below ---

For a bit more elaboration, I logged over 8000 packets associated with nationalrail.co.uk when searching for a split...

Very good point. I should make it much clearer that people will be searching direct againt the NR website. 8000 sounds like a lot, but I am aware of some inefficiencies that I could improve.

In its defence - it uses a single HTTP connection (all over a single TCP connection). It's not firing up multiple threads and completely hammering the NR website.

 

[yorkie]

It's flawed, but if you want feedback, see my PM.

It also doesn't factor in over distance.

Good - there would be so many issues with that, and it would result in many anomalies being withdrawn by ATOC!

 

[Nym]

Very good point. I should make it much clearer that people will be searching direct againt the NR website. 8000 sounds like a lot, but I am aware of some inefficiencies that I could improve.

In its defence - it uses a single HTTP connection (all over a single TCP connection). It's not firing up multiple threads and completely hammering the NR website.

Yes, but I invite you to look at a packet trace for what ACTUALLY happens, remember you're using org.apache.* libraries and that behave very differently to how you'd expect.

From what you've said there I'm assuming you have a connection object that you don't close, it doesn't make any difference to the actual packets sent if you do that, and if you look into what HTTP requests actually consist of, it's a shed tonne of data being thrown around, this you can't avoid, but if 100 people used this service at once for a complex unlimited split, it would probably put NR's website into cutdown mode, and they'd start searching for the software that makes IPs send thousands of packets at once.

PS: If you're wondering what qualifies me to be so critical, the report I'm currently writing is for one of the UK's top university's faculty of engineering, regarding bandwidth critical server/client http communication, just to give you a background into one of the areas I'm rather knowledgeable on.

 

[trainfanatic]

Sorry for not responding earlier....her inside needed some shopping done!

I tried to open the programme initially direct from the URL in your post...this failed saying no such programme or similar.

I then entered the URL directly and although the programme opened, there were a list of issues associated with it. Should have taken a scrren shot...

 

[WestCoast]

Looks interesting.

So, in layman's terms it's a type of screenscraper with local analysis? Screenscrapers have existed in the airline/ferry/package holidays ticketing world for a long time...

 

[Nym]

It's not the scraping thats the problem, it's the masses of requests...

And the fact it will probably mean split ticketing will either get made more difficult and/or fares change to make it pointless.

 

[RJ]

Just tried my favourite mainline where I can get a single at any time after 08:25 for £44 (Priv price *4) with splitting and it has come up with £86.50 as the cheapest price, even taking into account available off peak tickets.

It's actually an excellent bit of software with a very nice GUI and will undoubtedly save people money (until someone upstairs does something about it .) However, it's not a substitute for doing a good bit of homework, which I now have to do as my season ticket expires today . There are other things that need to be understood to make full use of splitting opportunities and has been mentioned, the less publicised they are, the less likely it is they will be removed.

 

[spasmj]

You may not at any time modify, store, copy (including for example screen scraping), extract, reutilise, distribute, transmit, display, perform, reproduce, publish, license, create derivative works from, transfer, or sell, distribute or create any information, products or services obtained from, linked to or using this Web Site and any data therein or that may provide users with the ability to do the same, unless approved by TISL through a signed licence agreement.

Its pretty clear that you aren't allowed to do this.

 

[ralphchadkirk]

You're not no, and as above, I don't think ATOC would be happy about that many enquires in one go. And btw, it's YOUR IP address making the enquires, so it's YOU that DBS/EWS (The people who run NR's servers) will be annoyed with...

Most railway system servers are managed by Atos Origin or Clyxx...

 

[Nym]

Most railway system servers are managed by Atos Origin or Clyxx...

NR's server's are kept by EWS though. Do a lookup on NR's IP addresses...

And spasmj, thanks for finding that, it's something I really couldn't be bothered reading...

 

[ralphchadkirk]

networkrailmediacentre.co.uk - Glide Technologies Ltd
networkrail.co.uk - Atos Origin
nationalrail.co.uk - ATOC Ltd
Unless there's something I'm missing here?