Day 23 - Indian Railways DOS based systems (?); A great trip to Calcutta

Today was a trip to Kolkata for Biswa Mast Aadmi. It was a a GREAT SHOW! And it was a wildly succesfuly trip also. Generally, my plans are foiled because of one thing or the other, today everything worked well. There’s a first!

On my way to Howrah, I had the chance to look at the screen on which the people behind the ticketing counters in the KGP Station find ticket availability between different stations, at different dates and different ticket classes (Sleeper, 3AC, 2AC, etc)

Ticketing is done on Text User Interfaces. Which apparently run using the VT220 Terminal. I discovered this from this incredibly unhelpful answer on Quora.


There are a few things to note here. The IRCTC apparently has a huge Database which is geographically separated, etc etc. That’s the backend part, that doesn’t matter for the present discussion.

The VT220 terminal itself was introduced in 1983. Yeah, 1983. That is about 8 years before Linus Torvalds started working on Linux. It’s also 34 years old now. Using a 34 year old frontend for ticketing? I can’t see any real reason for that to be an intelligent choice. Isn’t it just better to update atleast to a latest Linux version?

That aside, let’s look at how the DB is queried.

  1. Entering the date, origin station and destination station, and the ticket class: leads to a table that has trains running on that between the two stations and the availability for each of the trains.

  2. Almost immediately, if there are no tickets, or even if there are tickets, people want to try higher classes. This requires a change in one of the parameters.

  3. If higher classes are not available for that date, people who are willing to be flexible (which is almost everyone) want to know the availability of the earlier day, the next couple days, etc.

Achieving all of this in a crowded station, where noise levels are pretty high already leads to people having to shout to make themselves heard. It leads to chaos. There are 2 ways out of this:

  1. A kiosk at the station: People should ideally use the IRCTC website for checking availability. Or use the IRCTC mobile application. If they can’t, there should be GOOD kiosks (for some reason, all kiosks have very very bad touch screens and response times that range between 10s of seconds to 100s of seconds. Why not just use a keyboard and a mouse? Why this obsession with touch screens?)

    Sidenote: The IRCTC mobile app is the WORST MOBILE APP that has ever been written. When you have clicked on “Book Ticket” and gone through the payment and are waiting for your PNR, the application starts showing you a non-cancelable Android notification. The infinitely spinning ring. There’s no progress notification, you can know nothing. When you are on something like LTE (which I was once on, That’s where all this ranting is coming from), this takes forever. You can keep waiting, and it will eventually show you a message that’s a rough equivalent to “We couldn’t book the ticket. Money will be refunded in a few days. Try to book your ticket again”. Now, there are plenty of mobile applications out there with payment SDKs which are way way faster than whatever these people are using. I am sure that any company would be happy to provide their services to IRCTC atleast at a lower cost, not that cost would ever be a prohibiting factor for the leading long-distance transport provider in a country of 1.2 billion people. But seriously, GET YOUR ACT TOGETHER!

  2. Text interfaces are incredibly powerful. The terminal is one of the biggest examples of this. Skilled users can use text interfaces much faster than anything that requires touching or clicking on buttons with a mouse. That said, the data that comes back from the commands should also make sense. Take ls for example. ls is great, ls -ls is better when you want to find out the permissions of all the files, ls -lsh when you want to include hidden files, ls -lg when you want to see the group that owns the files and directories and so on. Basically, command line arguments. Would it be so hard to design a system that parsed something like:

    $ GET KGP CST 26 +-1 032
    # GET trains from KGP to CST for the 26th of the present month.
    # show availability for the next day and the previous day.
    # show availability for ticket classes 2nd Sleeper (0), 3AC (3) and 2AC (2)
    $ GET KGP CST 4n +3 32
    # KGP to CST on the 4th of the next month. Show avail for the next three days
    # as well. Show only for 3AC and 2AC

    Given enough time, writing a fast parser (preferably in C, coming from the benchmarks discussion in yesterday’s post) for these kinds of commands will not be hard.

There are some major caveats to this post. Not in the proposals that I have made, I beleive they are fine, definitely not rock solid because I know very little about ticketing except from whenever I have booked tickets.

  1. I don’t know if the TUI that they use is the VT220. That was from a Quora answer dated December 2014. Things shouldn’t have changed in this short a time.

  2. The backend that Indian Railways uses. It’s so hard to figure out what exactly they are using. There’s the website of the Centre for Railway Information Systems, there is a dedicated page to the Passenger Reservation System with a “Technology” page. It doesn’t mention if this is a project that’s being developed for later installment, or it’s already being used everywhere. It’s all a bit shady.

Finally, I have seen Indigo (the airlines) use a similar TUI interface. They use it for the generation of boarding passes and baggage tags at the check-in counter. The people working on them are highly skilled at using these terminals. So, clearly TUI has some advantage, whether IRCTC is tapping into all of it or not is questionable to me. I will continue to try to dig up data about this, but I get the feeling that that is going to be tough.

POST #23 is OVER