DEV Community

Cover image for Shell installers may be stronger than package installers on GitHub
Lucian Green
Lucian Green

Posted on


Shell installers may be stronger than package installers on GitHub

Scenario: I want to make installing my repository that is dependent on another of my repositories easier. I want to make the installation easier by doing it "in my way" (with a shell script) rather than package installers, which allows command-line installation of repositories, at the cost of development speed.

Wouldn't the idea be installing the repository with a particular version that corresponds to the version of the repository dependent on it? For example, an inductive algorithm finder of version 1.0 requires an interpreter of version 0.5. I saw the versions page somewhere while checking on the traffic to my GitHub repository.

I read on a well-known repository that SemVer is a preferred system of organising versions in the x.y.z format.

Is my repository ready for the industrial exposure of availability on the command-line? Is it just me, or did I add too many unnecessary, bug ridden features meaning I am grateful I kept an old version as a separate repository (some may have kept it as a separate version) and I can transfer the best (working) features from the new to the stable version... Why is my stable release my first of my three generations of my program - will it be too simple for be available on a command line argument?

Do I get credit for these uses of my software - I always wondered if people would use my programs when I had an Apple IIe, and was very steadfast about putting my name and the date in the documentation of the program. What if I make a famous program, does that make me famous?

Do I need a copywriter for my documentation page, you know that one that someone seems to copy onto various websites? Different in what way? Whatever happens, I don't want to be that person who wrote too much or too little documentation - "where is it?"

I could think of more to write - who (with which OS's, etc.) needs my program, do I need a single set of instructions on a page for using my dependency of repositories - with training, walkthroughs, even videos?

So, would using a package installer allow a user to search through a description of my repository and my algorithm given greater awareness? Should I label it "what I got up to" and "what I want, or rather how it fits into the eco-system of my other repositories"?

I just wish SWI-Prolog had a verify (input) feature, or should I put it in myself? Maybe I should explain how users should format input into my algorithm.

Photo by Bob Smith from FreeImages

Top comments (0)