System-level vs application-level package distribution
If I create a FOSS command-line utility (i.e. an application rather than a library) and I want to distribute it as a package for convenient installation, is it better practice to do this via:
- application-level package management (e.g. RubyGems, if it's written in Ruby; EasyInstall if Python, etc), or
- systems-level package management (e.g. MacPorts for the Mac, Debian Packages for Debian, etc).
NB. If you believe there is no generally correct practice but rather that there are particular circumstances under which one or the other approach is clearly best, then what are these circumstances?
It seems to me that the procedure followed in the case of applications whose users are happy with the way they're distributed, is roughly as follows:
Take, for example:
- Bundler: written in Ruby and useful to Ruby developers; distributed via application-level package.
- fwknop: written in Perl (originally, at least) and useful to end-users; distributed via systems-level packages/installers.
- Pandoc: written in Haskell and may be useful to Haskell developers in the course of developing Haskell applications, but also useful to others; distributed both via application-level and via systems-level packages.
There are probably better examples than these. Also, my conclusions so far are just observational ones: I haven't found a reference laying out this decision flow explicitly. As such, I'd welcome a more comprehensive answer.
NB. The code for the flowchart is online, in case you'd like to adapt it for your answer. To be able to use it, you'll need a working LaTeX installation.
- How to add a hover-over tooltip to rhandonstable header cell?
- How to detect the right encoding for read.csv?
- Rcpp Rf_warningcall compiler warnings
- Modify the name of factor variables in lm function(summary function)
- Extreme value analysis and quantile estimation using log Pearson type 3 (Pearson III) distribution - R vs Python
- How to hide NAs when using xlsx::saveWorkbook?
- How do I retrieve a simple numeric value from a named numeric vector in R?
- Adding an image in navbar and adjusting alignment in Rmarkdown
- Matching pair-wise columns from left to right across rows in one dataframe to another dataframe and adding new columns with matching values
- Income to outcome flow chart in Sankey plotly R
- color mapping in geom_conn_bundle not showing correctly
- Print R package startup message AFTER automatic package conflict messages instead of before
- Summing a set of R dataframe rows (column-wise), while retaining the first n columns
- Added variable / partial regression plots for groups in an interaction?
- how to make a topoplot in R with coordinates variable distribution
- List of all functions in base R?
- Plotting multiple plots for different initial conditions in one graph
- Printing repetitively on the same line in R
- Generating UI/Server based on initial selection
- Subset dataframe based on pickerInput
- How to let user pick the data in R-shiny?
- Couldn't show my simple bar charts separately on Shiny R dashboardBody
- How to programmatically filter contents of a second shiny app displayed via iframe
- How to select specific interesting groups for the boxplot in R Shiny app?
- Crosstable and Plot grouping with reactive values
- Is there a way to make multiple Shiny picker inputs where the selections must be disjoint?
- Delay/avoid duplication of shiny server side functions until after credentials
- Predictions only returns value "1"
- How to display a busy indicator in a shiny app?
- Append doesn't work when writing to CSV in R