The 100-Year Software Application | Hacker News Reader

The 100-Year Software Application

blog.smartbear.com

31 pointsBaustin32 comments

I have an acquaintance who works in the Gulf of Mexico anchoring platforms to the sea floor. They rate their anchors by the size of the storms they're engineered to withstand - a 10 year storm, a 100 year storm, etc. It's more difficult and correspondingly more expensive to place a "100 year storm" anchor. It's hard to argue with the reality that not all jobs require "100 year storm" anchors.

Arguing that software engineering is different is a result of inexperience. Not everything we build needs a "Rolls Royce" level of polish or a "Golden Gate Bridge" level of reliability nor is it a sustainable approach. Economic realities will drag your architecture astronauts back to Earth.

Spend time understanding the needs of your critical infrastructure and design for that.

The message I got from this article is that there is something deeply satisfying about doing "100 year storm" work, not than that it is economically feasible or wise. The reference to woodworking and how unprofitable it is seems to imply that it is inherently uneconomical.

Not everything we build needs "Rolls Royce" level of polish, but it might be nice to build a few things to that level of polish. Compared to the amount of code that is being written, vanishingly little of it, if any, is really being built to stand the test of time, even if it is being built to very high standards.

I'm the author of that post. And I certainly did not mean to imply that 100-year-furniture is not economical! It costs more in the short-term (for instance, the Stickley couch I bought was $6,000), but I know it'll last for decades, so I don't begrudge a cent of that expense. Especially since it's so gosh-durn beautiful and comfortable and does exactly what I want -- just as good software does.

I got my great grandma's couch reupholstered. The couch itself was free to me and I expect it to outlive me as well. I'm pretty sure she didn't get it new. I'm guessing manufactured in 1880 or so.

I dunno. It's pretty comfortable, and as you say, that couch problem is solved.

I'm sorry for both misunderstanding and misrepresenting your argument. I'll have to give a proper response some thought.

Thank you for such a gracious reply! In turn, I should apologize for writing something that could so easily be misinterpreted.

Oh hardly!

100 years is peanuts, relative to http://en.wikipedia.org/wiki/Maeslantkering, which was designed for a 10,000 year storm. Basically, the greater the damage if anything goes wrong, the sturdier things should be. This one has been in operation for 15 years, but has never been truly used yet.

I think a big part of this is that business requirements for software tend to change pretty quickly; where the requirements for physical things don't.

A desk from 1900 almost exactly fits the requirements for a desk in 2013 (other than what's in style, and the 'timeless' look often works). Even the 'foundational' software and reports in a business have to change due to the market and goals of the business evolving.

Antique desks don't have cable holes or filing drawers, but they're still desirable for their robustness and elegance. A similar statement could be made for UNIX utilities.

I'm pretty sure that emacs and vi(m) fit into this category, as do a number of unix command line tools (wc, cat, grep). Outside of that... heh... I hope most of the drek out there is replaced.

Eclipse exists since 2001/2002 and still one of the most popular IDE. I don't think Eclipse is going to go anywhere in the next 10 years.

Emacs/Vim definitely exist longer but Eclipse is joining the ranks.

Eclipse survival is an interesting question. I like it and use it all the time, but I kind of figure it is so integrated to its environment that it will go extinct quickly when another environment becomes dominant.

It will still float around for some time...

Thing about Eclipse is that it is cross-platform and have bajillion plugins.

Would you say the same about Visual Studio? Netbeans? These have had longer tenure than Eclipse.

What makes emacs/vim longer-lasting to me is their reach across platforms _and_ their simplicy of interface. When a new interface comes out, they just plug in to the old terminal emulator that's already stood the test of 30ish years...

You can say that for VS.NET and NetBeans.

Eclipse is also cross-platform and once you thought Java development dried up comes along Android. I've seen some C/C++ people in the embedded world to use Eclipse more and more lately.

Visual Studio, Netbeans, Eclipse, etc. get frequent updates. They get replaced on a regular & frequent basis with something similar. It's not "hundred year furniture", it's replacing the sofa every year with a new sofa featuring better upholstery, stronger joints, and better stain resistance. Might be the same general design, made by the same manufacturer, but it's not the same instance of "sofa".

It's the emacs/vim example which is more applicable. The core code hasn't changed for a very long time. The facade may change, but the operation behind that interface is unchanged.

Methinks the article refers more to deliver once, update never, restart never, crashing intolerable. Having a pacemaker makes one keenly aware of the difference: just a battery replacement (4-8 years) is a major undertaking, deliberate restarts are avoided and undertaken with grave care, and crashing, well, I wouldn't notice that problem for long.

I don't see Eclipse as a 100-year program in its present form. I think it is too tied to Java and the JVM; it's also terrifically tied into the corporate system of today.

When the jvm fades, I think eclipse will as well.

Check my other comments regarding embedded software development in C/C++.

There are tons of plugins even for Arduino, Lego, etc.

Eclipse is pretty quick to pivot. They've been getting better PHP ability. They're working on getting better Scala integration. Erlide exists now and doesn't suck. They're getting better with JS. And for most of these there are competing teams trying to solve the same problem better/faster.

Java as the base seems to help by making it cross platform. So I can learn it in Windows and transition to a client with Linux and be minimally impacted.

Now what will make Eclipse go down is another Juno. 100 year software has to be solid and performent. Juno bungled this. They've gotten better, but it's still a black eye.

I think there are a few banking mainframe apps that are 30+ years old...

Only 30? I'm not-maintaining a 20-year old app at the moment (that is, I should be maintaining it :P ) and it's pretty new by industry standards. There are some local apps that are 40 years old (I think there's some University software that was programmed in the late 60s), so I guess there must be 50-year old apps somewhere else :)

Really, that's quite likely, but I don't actually know of any myself, so I chose to keep it to software that I (1) knew about and (2) others did as well.

It's worth pointing out (because I'm a Lisp nerd) that there is a decent amount of Lisp code that is still executable today that was written a long time ago. In terms of software systems, https://en.wikipedia.org/wiki/Maxima_%28software%29 is the one I know about; there might be quite a few others I don't know about.

Also, I've gotten the impression that there is some sort of slow move to Java/.NET from the COBOL systems.

There are a few source files in one of our applications[1] that have last modified dates back in 1994.

1. Mostly C application that's had >$1bn in sales over the years, but it's been in active development/enhancement since 1992. My oldest contribution to it is probably 1997.

Certain types of solutions are more likely to last. For example an iPhone app is not likely to still be used in 100 years but say a core compression algorithm could.

Basically UI centric solutions have a shorter span on average than Algorithms to complex problems.

Most serious woodworkers talk about creating "100-year furniture." You won't get rich as a custom woodworker, but you can be reasonably sure that the table you built will be used by the grandchildren of the client for whom you build it today.

I read someone's comment on IKEA furniture recently: "you are buying furniture that you and your descendants will use for hundreds of days"

I would guess it varies with model. The Poang chairs are afaict flagship products. Mine are about a decade old now, but replaced the cushions once. I have to retighten the bolts every couple of years. I can imagine kids breaking them from being dumb, but from ordinary (daily) use, they don't seem any worse for wear.

I busted a Poäng chair once - the wood of the vertical pieces closest to the floor split and I fell backwards.

However, I think was quite a few years old and a floor model, so who knows what abuse it was subjected to in the store.

Supposedly, Ingvar Kamprad has a 30+ year old Poäng.

Businesses are more like software than furniture are. The Japanese hotel, Hoshi Ryokan, that is still run by the same family over 1300 years must have some really good "software".

Here's the list of oldest companies to get inspiration from: http://en.wikipedia.org/wiki/List_of_oldest_companies

I was using troff yesterday. The technology dates back to '60s and is still widely used.

Dan Bricklin wrote a very nice article on the subject: http://changethis.com/manifesto/show/6.200YearSoftware

Reply on news.ycombinator.com