4 ms·

> I see this sentiment a lot, but it’s not a useful one... Consider this response for entertainment purposes only. It’s easy to become insult(ed|ing) or get in

• effie 9y ago
  > sh(1) (or bash - nobody cares) features feel archaic and brittle Maybe some shell features are not that great, but in general I feel the opposite. Many of the shell features are great and lacking in other, more general-purpose languages - for example, program composition via pipes, stdout/stderr redirection syntax, direct access to the filesystem, interchangeability of a function and a program call, and others. When somebody decides to write a big system, like Kubernetes, they better be pragmatic rather than idealistic. That means choosing the right tool for the job, even if this tool has some deficiences or is considered quirky by some people. Manipulating VMs,containers on localhost or remote hosts, interacting with the OS and the filesystem, all on a unix-like system, it seems to me like shell is the best tool there is for most of the task.
  • bch 9y ago
    > composition via pipes That’s a good, big one in favour of shells. Re: file system, std err/out, I’m not convinced, but pipes alone are worth a lot. I wasn’t thinking at all of pipes, and you’ve tipped the scales, but is it enough? Do Tcl/Perl/ruby/python/... compete in that space?
    • jancsika 9y ago
      > Re: file system, std err/out, I’m not convinced, but pipes alone are worth a lot. How does the existence or lack of a standard data interchange format affect the value of those pipes?
      • bch 9y ago
        It seems to me that theoretically the effect is brutal. Practically, it’s negligible. We pipe everyday, all day, and the world carries on, so that’s a pragmatic vote of confidence for pipes. Are you trying to get me to write a “worse is better” essay that ends with “...and the existence and continued durability of shell scripts is proof enough that their place is not only justified, it’s vital!”, because it won’t work. Pipes are good. Shell scripts should be looked at suspiciously if they’re more than 20 lines long. ;)
        • philipov 9y ago
          Pipes aren't even something shell scripts have a monopoly on. They're just as easy to use in Perl. Perl may not be known for its readability, but it still ends up miles ahead of an equivalent script in bash. I do wish there was a first class syntax for it in Python (but you can get most of the way there by using generator/coroutine pipelines).
    • chubot 9y ago
      FWIW this post is related: Pipelines Support Vectorized, Point-Free, and Imperative Style http://www.oilshell.org/blog/2017/01/15.html http://www.oilshell.org/blog/2017/01/15.html I started writing a series on "what shell can do that other languages can't", but I put it on hold until the shell is in better shape. I hope to resume that series in the next few months.
• chubot 9y ago
  It's clear from these comments that I need to explain the project more concisely on the blog (which I mentioned in the conclusion), but here is a short response: (1) In retrospect, I should have responded to the first commenter differently. He was clearly angered by having to debug other people's shell scripts in the past. But that is exactly what I'm trying to fix -- I'm trying to provide an UPGRADE PATH OUT OF BASH. I explicitly state that in this long post, which I think some people have missed. These posts show one part of the plan: http://www.oilshell.org/blog/tags.html?tag=osh-to-oil#osh-to-oil http://www.oilshell.org/blog/tags.html?tag=osh-to-oil#osh-to... (2) I'll repeat my objection: trying to convince people not to use bash is going to be about as successful as convincing people not to use PHP. Not only is it not a solution to the problem, it also ignores the humongous installed base of PHP code, like Wikipedia, etc. Even if another line of bash never gets written, you'll still have to deal with it on a regular basis! I am providing a path out of bash, while other people are just wishing the problem would go away. The world is not how you want it to be, but exhortation on online forums isn't going to change the world. Less charitably, "you're wasting your breath". I need to write a blog post entitled "Reimplementing Bash is the Only Way to Kill It". This is analogous to how Facebook is "removing" PHP from their codebase by developing the Hack language and VM. It's also analogous to how Microsoft "killed" Lotus and other competitors by understanding their file formats. (3) I also don't think you understand why cloud projects like Kubernetes are using shell scripts, probably because you don't work in that domain. Shell really is the best tool for that job. I think it's presumptuous to state that the people writing that code don't know what they're doing. A lot of the project is in Go, but a lot of it is in shell, and there's probably a reason for that other than ignorance.
  • bch 9y ago
    To start, I think your project is excellent and I don't mean to take away from your work. It has bore fruit, which is an accomplishment on it's own, to say nothing of the intrinsic value of simply investigating and hacking in whatever domain interests you. I like to think I'm fond of hacking in "unsexy" spaces, and I think hot-rodding a shell is that space. Re: 3) above -- Curious to hear about the reasons. Re: "...people writing that code ...other than ignorance" -- I'm not trying to imply people working on this are idiots... though w/ broad-sounding statements like "scripts with more than 20 lines should be looked at suspiciously" I can see how someone might read that. I think there are myriad poor reasons an shell script might be inappropriately used. There's also a case for "if it's not mine, who am I to judge?". I'd be curious to hear points of why shell is especially suited for (e.g.) cloud projects like Kubernetes, though. Good luck with your work.