5 ms·

Speaking as someone who lived through this period during the formative years of my programming career, the author has completely missed the most influential lan

by sulam 12y ago

Speaking as someone who lived through this period during the formative years of my programming career, the author has completely missed the most influential language of the time and, in my mind at least, the number reason why Perl lost. Yes, I'm talking about Java. It's easy to hate now, but Java back then replaced all the server-side Perl programming that I did in the space of about 3 years, from 1994 where CGI programming was king to 1995 and the applets craze to 1999 when I was still regularly having conversations with people about the benefits of Java vs Perl to 2001 when you were actively hurting your career by not learning Java (not least because it seemed like the only people hiring were doing enterprise Java with horrible things like EJB and CMP/BMP). Why did Java beat Perl? Well, there's a lot to that -- but at least part of it has to do with that fact that Java was simpler and didn't have nearly as many tricks up its sleeves (also known as the write-only problem). This is similar to the argument people today make when picking Go over other options like Scala. I also don't think it hurt that Java came out of Sun, which on the one hand was extremely influential because they sold the hardware everyone with money used, but on the other wasn't influential at all because they were "big iron" to the web's "why would I buy a Sparcstation when I can just put a PC under my desk?" Fundamentally, though, Java was very successful at becoming the language you wrote code in if you wanted to be taken seriously as a software engineer building web applications and you didn't already have 5+ years of C++ experience. Perl programmers didn't get the same respect, and so Perl died.

nemo 12y ago
There were two sides - PHP was eating a ton of share from Perl for people and businesses who just wanted to get some simple web thing going. Sun pulled out the "Java is Enterprise" card early and successfully marketed it as being a professional scalable web programming solution for enterprise class applications with a few targeted buzzwords for whatever market segment they were going for. Perl seemed like it was always looked at as an amateur tool (much as PHP/ASP generally were), while Java was taken seriously in "enterprise" spaces (largely due to Sun's success at marketing Java to businesses). There's a wrinkle, too. Perl isn't just a web thing. It was (and is) still used in "enterprise" spaces for scripting on servers. It's still out there being used to create scripts rather than applications. In that space Ruby/Python are competing, but there's still plenty of sysadmins who use Perl since it works.
- pjmlp 12y ago
  > In that space Ruby/Python are competing, but there's still plenty of sysadmins who use Perl since it works. I don't know how it looks now, but if you login into the likes of HP-UX, Aix, Solaris and so on, usually only Perl is available. This is what made me learn Perl, back when I was mostly into Python (.com days).
bad_user 12y ago
I've dabbled in the source-code of many projects. And personally I've never seen projects more incomprehensible than some of the ones I've seen built with Java. Of course, I've also seen well written Java code, however Java projects tend to be massively over-engineered and tasteless. Its popularity doesn't help either, the signal to noise ratio in the pool of Java developers being awful. On people picking up Go over other options, well that's because people never learn and we are bound to propel shitty languages ad-nauseam.
- claystu 12y ago
  What do you mean about Go? (And no...I'm not trying to pick a fight--just curious)
  - kasey_junk 12y ago
    There is a common theme to early Java discussions and discussions about golang right now. I'll sum it up as "All that power/functionality/expressivness of language x is really a distraction from getting stuff done. Further it complicates the software development model of large disparate groups." I'm not sure that theme is wrong even, but to many of us who lived through early java, golang's least common denominator approach to developer sourcing is eerily familiar.
    - jerf 12y ago
      I agree and have made that point before. Java tried to be a language for "enterprise programming", which tried to keep the language so simple that having lots of people wouldn't muck up your code. IMHO, it failed epically at this, and today we would basically laugh at the idea that Java is good at that, but it was in fact a design criterion back in the day. One could make the case that it sort of succeeded at the isolation bit, but at the cost of making everything so verbose that it's absurd and a constant stream of technologies to deal with this has been coming out for Java ever since. Despite the taint Java left on it, it's actually a good idea! It's a common use case. After years of programming in Perl in a fairly large programmer environment I can see the virtue of the idea of using a simpler language helping keep a lot of programmers in sync in their code base, instead of living in a world where I can look at the source and virtually instantly tell both when and who it came from due to massive differences in dialect. But Java seems to have failed so badly that it soured everyone on the entire idea, and nobody has tried since. I think Go actually has a better chance at this, not for any large reason but for a series of smaller ones. Defaulting to object composition, once you understand it, is a powerful-yet-simpler approach to putting objects together than inheritance, and makes it much easier to both use the objects of others and to create objects to use for others, because you don't need to worry about what will be F'ed up in the "child classes", as there is no such thing. Implicitly satisfied interfaces don't sound like they'd be a big change over Java, in practice it's night-and-day to be able to declare interfaces that are automatically satisfied by objects you don't control. Contrast how many ways there are to write the same Go function as there is to write it in, say, Scala. And writing some decent multithreading primitives into the core is probably a good idea for this use case, too. It won't happen tomorrow or next year, but, arguably, rather than Rust or even Python, the language that Go is the biggest threat to in the long term is Java. And it's a threat for precisely the reasons that people on HN tend to complain about. Indeed, I personally have the same objections that most HN'ers do for personal usage, but right now bar none Go is the language I'd most like to work in, precisely because it was written for the Google use case which also happens to be the one I have.
nailer 12y ago
Ditto on the Sys Admin side with Python - Perl had modules for everything, but the inbuilt punctuation variables mixed with regexs for everything and the flexibility Perl offered led to a perception that nobody could maintain anybody else's stuff. Sys admins were still about getting shit done quickly at scale, and Python slowly began to replace Perl. Python 2 had string methods built in, Red Hat switched everything to Python, tool.s like LXML started coming out right when we all had to munge XML all day and suddenly that was it.

by sulam 12y ago

nemo 12y ago

There were two sides - PHP was eating a ton of share from Perl for people and businesses who just wanted to get some simple web thing going. Sun pulled out the "Java is Enterprise" card early and successfully marketed it as being a professional scalable web programming solution for enterprise class applications with a few targeted buzzwords for whatever market segment they were going for. Perl seemed like it was always looked at as an amateur tool (much as PHP/ASP generally were), while Java was taken seriously in "enterprise" spaces (largely due to Sun's success at marketing Java to businesses). There's a wrinkle, too. Perl isn't just a web thing. It was (and is) still used in "enterprise" spaces for scripting on servers. It's still out there being used to create scripts rather than applications. In that space Ruby/Python are competing, but there's still plenty of sysadmins who use Perl since it works.

pjmlp 12y ago
> In that space Ruby/Python are competing, but there's still plenty of sysadmins who use Perl since it works. I don't know how it looks now, but if you login into the likes of HP-UX, Aix, Solaris and so on, usually only Perl is available. This is what made me learn Perl, back when I was mostly into Python (.com days).

bad_user 12y ago

I've dabbled in the source-code of many projects. And personally I've never seen projects more incomprehensible than some of the ones I've seen built with Java. Of course, I've also seen well written Java code, however Java projects tend to be massively over-engineered and tasteless. Its popularity doesn't help either, the signal to noise ratio in the pool of Java developers being awful. On people picking up Go over other options, well that's because people never learn and we are bound to propel shitty languages ad-nauseam.

claystu 12y ago
What do you mean about Go? (And no...I'm not trying to pick a fight--just curious)
- kasey_junk 12y ago
  There is a common theme to early Java discussions and discussions about golang right now. I'll sum it up as "All that power/functionality/expressivness of language x is really a distraction from getting stuff done. Further it complicates the software development model of large disparate groups." I'm not sure that theme is wrong even, but to many of us who lived through early java, golang's least common denominator approach to developer sourcing is eerily familiar.
  - jerf 12y ago
    I agree and have made that point before. Java tried to be a language for "enterprise programming", which tried to keep the language so simple that having lots of people wouldn't muck up your code. IMHO, it failed epically at this, and today we would basically laugh at the idea that Java is good at that, but it was in fact a design criterion back in the day. One could make the case that it sort of succeeded at the isolation bit, but at the cost of making everything so verbose that it's absurd and a constant stream of technologies to deal with this has been coming out for Java ever since. Despite the taint Java left on it, it's actually a good idea! It's a common use case. After years of programming in Perl in a fairly large programmer environment I can see the virtue of the idea of using a simpler language helping keep a lot of programmers in sync in their code base, instead of living in a world where I can look at the source and virtually instantly tell both when and who it came from due to massive differences in dialect. But Java seems to have failed so badly that it soured everyone on the entire idea, and nobody has tried since. I think Go actually has a better chance at this, not for any large reason but for a series of smaller ones. Defaulting to object composition, once you understand it, is a powerful-yet-simpler approach to putting objects together than inheritance, and makes it much easier to both use the objects of others and to create objects to use for others, because you don't need to worry about what will be F'ed up in the "child classes", as there is no such thing. Implicitly satisfied interfaces don't sound like they'd be a big change over Java, in practice it's night-and-day to be able to declare interfaces that are automatically satisfied by objects you don't control. Contrast how many ways there are to write the same Go function as there is to write it in, say, Scala. And writing some decent multithreading primitives into the core is probably a good idea for this use case, too. It won't happen tomorrow or next year, but, arguably, rather than Rust or even Python, the language that Go is the biggest threat to in the long term is Java. And it's a threat for precisely the reasons that people on HN tend to complain about. Indeed, I personally have the same objections that most HN'ers do for personal usage, but right now bar none Go is the language I'd most like to work in, precisely because it was written for the Google use case which also happens to be the one I have.

nailer 12y ago

Ditto on the Sys Admin side with Python - Perl had modules for everything, but the inbuilt punctuation variables mixed with regexs for everything and the flexibility Perl offered led to a perception that nobody could maintain anybody else's stuff. Sys admins were still about getting shit done quickly at scale, and Python slowly began to replace Perl. Python 2 had string methods built in, Red Hat switched everything to Python, tool.s like LXML started coming out right when we all had to munge XML all day and suddenly that was it.