9 ms·

One thing I don't see mentioned often is that productivity in software development has skyrocketed, but it's hard to appreciate. Today's services are much more

One thing I don't see mentioned often is that productivity in software development has skyrocketed, but it's hard to appreciate. Today's services are much more sophisticated and complex than just a decade ago. Even if some of that complexity is incidental and undesirable, it's undeniable that today a small team can pump out a much more refined app that entire companies did back then. This isn't only guess work, here is a real example: when I joined my first shop we had 2 sysadmin working full time to maintain a couple dozen servers. Over time they started introducing virtualization, automation, devops... by the end of my 7 year tenure, we had 250+ servers and only one sysadmin part time working on them. That's between 1 and 2 orders of magnitude improvement, and my company wasn't special by any means.

marginalia_nu 4y ago
You could also less charitably read that "growth" as using 250 servers to do the same job you previously did with two. I guess that's growth for sever manufacturers, but I'd expect with more powerful hardware we'd need fewer servers, rather than more.
- adgjlsfhk1 4y ago
  what you're missing is customer numbers. If you go from 20 servers to 200, and have 100x the customers, that's a massive win.
- angarg12 4y ago
  I'm not sure what part of my comment made you draw that conclusion. This was a science team and we used servers mostly for scientific computations. Sysadmins jobs was to operate and maintain the servers e.g. make jobs run, upgrade, patch, fix, etc. Virtualization massively simplified this, and automation decreased human intervention. With 250 we were definitely running bigger tasks faster e.g. a project that required 3 months of computing now takes 2 weeks instead.
SoftTalker 4y ago
Yes, but a person can write a paper just as well in Wordperfect 5 as the latest Google Doc or Office365 Word. How many millions of dev-hours have been expended on stuff that fundamentally doesn't do anything new? Edit: saying they don't do anything new isn't right. They do, especially with regards to collaboration. What I really mean is that they don't make the essential task -- the actual writing -- much easier. That's still human/mental work. Once you have come up with the words, putting the words into a document is the easy part.
- KronisLV 4y ago
  > Saying they don't do anything new isn't right. Perhaps it's possible that some people simply don't see the benefit of the newly added functionality, because they don't really use it? For example, personally I'd be served perfectly well with something like LibreOffice, so from my point of view, all of these new solutions do not indeed do much new (that I'd benefit from) and sometimes do things worse (cloud based platform with all the drawbacks that come with it vs local software in open formats). That's not to say that the point that you're making wouldn't be important, quite on the contrary - and even if the software doesn't do anything new, even in those cases one can make an argument for refinement being a worthy pursuit (as long as it's not a useless rewrite). Perhaps text editing isn't the best example, given how different people's expectations can be in that regard.
- wolverine876 4y ago
  > a person can write a paper just as well in Wordperfect 5 as the latest Google Doc or Office365 Word That seems hard to believe; try printing, or formatting. It's like saying 'a person can fly just as well in a 1930s biplane or a current 737.' Sure, they both fly. On the other hand, am I more productive in 2022 Emacs than I would be in 1985 Emacs?
  - dredmorbius 4y ago
    As Robert Gordon notes in his The Rise and Fall of American Growht, an air traveller in 1936 headed from New York to Chicago, could have made the trip in 3h 55m gate-to-gate, vs. 2h 35m today. But in 1936, arrival at the airport need only be 10 minutes prior to departure, gates were immediately next to the terminal entrance, and taxis or a chauffeured Cadillac limousine were 5 minutes from the gate --- the latter cost $10.50 in 2014 prices. The 2014 traveller must arrive an hour prior to departure, it takes 20 minutes merely to get off the plane on average (from the middle of the plane) on arrival, and 5--10 minutes to get to the taxi line. The net trip is only 20 minutes shorter than in 1936, despite the faster in-flight leg. Much more of that time is spent standing, waiting, or walking. Robert J. Gordon, The Rise and Fall of American Growth (2015), pp. 396-7.
kareemsabri 4y ago
I'm genuinely curious about this, and would like to see some data. When I started coding professionally the LAMP stack was the go-to, and your program would generally just do all the "work" in the context of a web request and then render an HTML page in response. Now, the systems are much more complex for sure (queues and asynchronous processing, event buses, microservices, multiple databases for different purposes) but I'm not sure that (functionality provided to business) / (number of developers) has increased all that much. Of course I could be dealing with a bias because as a novice programmer I worked on simpler programs and now I work on more complex / powerful programs.
- nine_k 4y ago
  Look at: Google Maps, Figma, any advanced enough web mail, Office 360. Think about implementing these using the LAMP stack + Javascript from 2000.
  - kareemsabri 4y ago
    Isn't that just saying browsers have gotten better? OP was talking about overall system complexity.
  - acheron 4y ago
    The growth of JavaScript is not something to be happy about.
    - alephnan 4y ago
      Why not?
- root_axis 4y ago
  > Of course I could be dealing with a bias because as a novice programmer I worked on simpler programs and now I work on more complex / powerful programs. The software techniques you've listed are among the reasons why it's even possible to build more complex and powerful programs. Forward progress in software is always resisted by the contemporary programmers of the day who are deeply fond of their chosen anachronism and don't like it when they perceive the feeling that it is threatened by a newcomer. This is an ancient law of software that few seem to accept, but c is to assembly as python is to c and javascript is to python. Software developers tend to be thoughtful people, these tools exist for a reason.
tluyben2 4y ago
We had one frontend engineer in the 80s/90s doing TP and Delphi frontends, and he was also doing some backend work because frontend work was very easy and fast. We now have 10 frontend devs and I don’t find it an improvement. It looks somewhat better sometimes, but it’s slower, more inconsistent (you can jump from 1 of the 100s of applications we did in Delphi to a random other and immediately understand it all if you use Windows) and incredibly costly compared.
alaricus 4y ago
Today's services are indeed much more sophisticated, but not necessarily better. This is because better and cheaper hardware made people sloppy, not because we have much better services.
- Gigachad 4y ago
  We do have much better services though. The feature set and scale of modern platforms and tools is mind blowing and so far ahead of what existed 10 years ago. If you released any of the major apps we used 10 years ago today, most people would consider them unusable garbage.
  - alaricus 4y ago
    I hate modern apps. They are bloated, slow, buggy and not at all customizable. Something went wrong in the last 10 years.

marginalia_nu 4y ago

You could also less charitably read that "growth" as using 250 servers to do the same job you previously did with two. I guess that's growth for sever manufacturers, but I'd expect with more powerful hardware we'd need fewer servers, rather than more.

adgjlsfhk1 4y ago
what you're missing is customer numbers. If you go from 20 servers to 200, and have 100x the customers, that's a massive win.
angarg12 4y ago
I'm not sure what part of my comment made you draw that conclusion. This was a science team and we used servers mostly for scientific computations. Sysadmins jobs was to operate and maintain the servers e.g. make jobs run, upgrade, patch, fix, etc. Virtualization massively simplified this, and automation decreased human intervention. With 250 we were definitely running bigger tasks faster e.g. a project that required 3 months of computing now takes 2 weeks instead.

SoftTalker 4y ago

Yes, but a person can write a paper just as well in Wordperfect 5 as the latest Google Doc or Office365 Word. How many millions of dev-hours have been expended on stuff that fundamentally doesn't do anything new? Edit: saying they don't do anything new isn't right. They do, especially with regards to collaboration. What I really mean is that they don't make the essential task -- the actual writing -- much easier. That's still human/mental work. Once you have come up with the words, putting the words into a document is the easy part.

KronisLV 4y ago
> Saying they don't do anything new isn't right. Perhaps it's possible that some people simply don't see the benefit of the newly added functionality, because they don't really use it? For example, personally I'd be served perfectly well with something like LibreOffice, so from my point of view, all of these new solutions do not indeed do much new (that I'd benefit from) and sometimes do things worse (cloud based platform with all the drawbacks that come with it vs local software in open formats). That's not to say that the point that you're making wouldn't be important, quite on the contrary - and even if the software doesn't do anything new, even in those cases one can make an argument for refinement being a worthy pursuit (as long as it's not a useless rewrite). Perhaps text editing isn't the best example, given how different people's expectations can be in that regard.
wolverine876 4y ago
> a person can write a paper just as well in Wordperfect 5 as the latest Google Doc or Office365 Word That seems hard to believe; try printing, or formatting. It's like saying 'a person can fly just as well in a 1930s biplane or a current 737.' Sure, they both fly. On the other hand, am I more productive in 2022 Emacs than I would be in 1985 Emacs?
- dredmorbius 4y ago
  As Robert Gordon notes in his The Rise and Fall of American Growht, an air traveller in 1936 headed from New York to Chicago, could have made the trip in 3h 55m gate-to-gate, vs. 2h 35m today. But in 1936, arrival at the airport need only be 10 minutes prior to departure, gates were immediately next to the terminal entrance, and taxis or a chauffeured Cadillac limousine were 5 minutes from the gate --- the latter cost $10.50 in 2014 prices. The 2014 traveller must arrive an hour prior to departure, it takes 20 minutes merely to get off the plane on average (from the middle of the plane) on arrival, and 5--10 minutes to get to the taxi line. The net trip is only 20 minutes shorter than in 1936, despite the faster in-flight leg. Much more of that time is spent standing, waiting, or walking. Robert J. Gordon, The Rise and Fall of American Growth (2015), pp. 396-7.

kareemsabri 4y ago

I'm genuinely curious about this, and would like to see some data. When I started coding professionally the LAMP stack was the go-to, and your program would generally just do all the "work" in the context of a web request and then render an HTML page in response. Now, the systems are much more complex for sure (queues and asynchronous processing, event buses, microservices, multiple databases for different purposes) but I'm not sure that (functionality provided to business) / (number of developers) has increased all that much. Of course I could be dealing with a bias because as a novice programmer I worked on simpler programs and now I work on more complex / powerful programs.

nine_k 4y ago
Look at: Google Maps, Figma, any advanced enough web mail, Office 360. Think about implementing these using the LAMP stack + Javascript from 2000.
- kareemsabri 4y ago
  Isn't that just saying browsers have gotten better? OP was talking about overall system complexity.
- acheron 4y ago
  The growth of JavaScript is not something to be happy about.
  - alephnan 4y ago
    Why not?
root_axis 4y ago
> Of course I could be dealing with a bias because as a novice programmer I worked on simpler programs and now I work on more complex / powerful programs. The software techniques you've listed are among the reasons why it's even possible to build more complex and powerful programs. Forward progress in software is always resisted by the contemporary programmers of the day who are deeply fond of their chosen anachronism and don't like it when they perceive the feeling that it is threatened by a newcomer. This is an ancient law of software that few seem to accept, but c is to assembly as python is to c and javascript is to python. Software developers tend to be thoughtful people, these tools exist for a reason.

tluyben2 4y ago

We had one frontend engineer in the 80s/90s doing TP and Delphi frontends, and he was also doing some backend work because frontend work was very easy and fast. We now have 10 frontend devs and I don’t find it an improvement. It looks somewhat better sometimes, but it’s slower, more inconsistent (you can jump from 1 of the 100s of applications we did in Delphi to a random other and immediately understand it all if you use Windows) and incredibly costly compared.

alaricus 4y ago

Today's services are indeed much more sophisticated, but not necessarily better. This is because better and cheaper hardware made people sloppy, not because we have much better services.

Gigachad 4y ago
We do have much better services though. The feature set and scale of modern platforms and tools is mind blowing and so far ahead of what existed 10 years ago. If you released any of the major apps we used 10 years ago today, most people would consider them unusable garbage.
- alaricus 4y ago
  I hate modern apps. They are bloated, slow, buggy and not at all customizable. Something went wrong in the last 10 years.