top | item 32023887

Ask HN: Is rebuilding an entire legacy software system stupid?

2 points| bveiseh | 3 years ago

I don't mean one legacy software, I'm talking a large set of intertwined and complex software systems that were build 30 years ago and still (inefficiently) in use today.

Is rebuilding it into one modern platform a stupid idea? And would a VC fund it?

7 comments

h2odragon|3 years ago

I'm sure theres stories of "we re-wrote everything in $(modern language) on time, in budget, and now everything works so much better"... I've never heard one tho.

If you want someone else to fund such a thing, it's probably gonna need to be something that has immediate, slam dunk sales prospects at then end... like "we're gonna re-write the post office systems and we know we can sell it because the Postmaster General is my father"

bveiseh|3 years ago

you're not far off, tho nepotism wouldn't be a contributing factor lol, see my commend below

mikewarot|3 years ago

>still (inefficiently) in use today.

What's inefficient about it? Computers are millions of times faster than they were, you're not worried about CPU cycles are you?

Is there a business need that the software can't meet? Or, is it just the general hassle of dealing with legacy code?

bveiseh|3 years ago

Not cpu cycles, I mean real system architecture inefficiencies + adding modern processes to improve performance. see my comment below on what I had in mind, but basically airline systems

akagusu|3 years ago

It's difficult to say anything without context.

What does it do?

What is the motivation behind the rewrite?

bveiseh|3 years ago

I want to re-write the systems that manage airline bookings, ticketing, and flight+crew management.

Airlines are always facing issues with ticketing and mismanagement of assigning air crews and pilots.. introducing a modern ML accelerated flow to determine how to balance ticketing with crews for flights, taking into account historical numbers of passengers for that time/route could have an enormous impact on airline efficiency.

LouisSayers|3 years ago

If you're asking yourself if it's a stupid idea, it sounds like you already know the answer...

Maybe if you could attack just a part of the problem, extract it, then you might have something?