All data must be under the same license

[krlmlr]

Step 1: find out if we can remove graphs without breaking downstream.

https://github.com/search?q=org%3Acran+%28repo%3Acran%2FclustAnalytics+OR+repo%3Acran%2Fdodgr+OR+repo%3Acran%2Fhydra+OR+repo%3Acran%2Fig.degree.betweenness+OR+repo%3Acran%2Figraph+OR+repo%3Acran%2Finterplex+OR+repo%3Acran%2Finvertiforms+OR+repo%3Acran%2Fmultinets+OR+repo%3Acran%2Fnetplot+OR+repo%3Acran%2FrSpectral+OR+repo%3Acran%2Fsand+OR+repo%3Acran%2FScorePlus+OR+repo%3Acran%2Fvsp%29+%2Fenron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs%2F+&type=code

R code to construct the query:

# Load necessary package
library(tools)

# Define the datasets from igraphdata
igraph_datasets <- c("enron", "karate", "yeast", "ukfaculty", "kite",
                     "macaco", "koenigsberg", "usairports", "dolphins",
                     "immuno", "foodweb", "polblogs")

# Identify reverse dependencies of igraphdata
igraphdata_dependencies <- package_dependencies("igraphdata", reverse = TRUE, which = "all")
igraphdata_packages <- unique(unlist(igraphdata_dependencies))

# Construct the GitHub search query using regex search
dataset_query <- paste0("/", paste(igraph_datasets, collapse = "|"), "/")
repo_query <- paste(paste0("repo:cran/", igraphdata_packages), collapse = " OR ")

github_query <- paste0("org:cran (", repo_query, ") ", dataset_query)

# Print the query
cat("GitHub Search Query:\n", github_query, "\n")

# Check for length limit
max_query_length <- 2048  # GitHub's max URL length
if (nchar(github_query) > max_query_length) {
  cat("\nWARNING: Query exceeds GitHub's search limit. Consider splitting it.\n")
}

Step 2: Evaluate next steps

TBD.

[krlmlr]

One search per package:

• https://github.com/search?q=org:cran%20(repo:cran/clustAnalytics)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/dodgr)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/hydra)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/ig.degree.betweenness)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/igraph)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/interplex)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/invertiforms)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/multinets)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/netplot)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/rSpectral)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/sand)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/ScorePlus)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/
• https://github.com/search?q=org:cran%20(repo:cran/vsp)%20/enron%7Ckarate%7Cyeast%7Cukfaculty%7Ckite%7Cmacaco%7Ckoenigsberg%7Cusairports%7Cdolphins%7Cimmuno%7Cfoodweb%7Cpolblogs/

Script:

# Load necessary package
library(tools)

# Define the datasets from igraphdata
igraph_datasets <- c("enron", "karate", "yeast", "ukfaculty", "kite",
                     "macaco", "koenigsberg", "usairports", "dolphins",
                     "immuno", "foodweb", "polblogs")

# Identify all reverse dependencies of igraphdata
igraphdata_dependencies <- package_dependencies("igraphdata", reverse = TRUE, which = "all")
igraphdata_packages <- unique(unlist(igraphdata_dependencies))

# Construct dataset search query using regex
dataset_query <- paste0("/", paste(igraph_datasets, collapse = "|"), "/")

# Construct GitHub search URLs
github_search_urls <- sapply(igraphdata_packages, function(repo_group) {
  repo_query <- paste(paste0("repo:cran/", repo_group), collapse = " OR ")
  search_query <- paste0("org:cran (", repo_query, ") ", dataset_query)
  paste0("https://github.com/search?q=", URLencode(search_query))
})

# Print GitHub search URLs
cat("GitHub Search URLs:\n", paste0("- ", github_search_urls, collapse = "\n"), "\n")

[krlmlr]

Only karate, UKfaculty (2-3x), yeast (1x) and enron (1x) appear to be used.

[krlmlr]

We can remove rfid, it's the only one with the contagious GPL-3 license, and license the package with CC-BY-SA.

Downstream: neylsoncrepalde/multinets#4 .

[szhorvat]

Only karate, UKfaculty (2-3x), yeast (1x) and enron (1x) appear to be used.

This is not a good way to test if these are used.

I like to think of "who are our users?" and "what functionality is used?" as consisting of two parts:

1. dependent packages, which are easy to test (CRAN even requires it) and easy to check what they use
2. academics who use igraph for their work (data exploration, research, teaching), but do not publish packages

I believe that group 2 is at least as important as group 1. Group 2 is the first group that was there, they are the main users and who made igraph popular. It has already happened that we assumed that a certain piece of functionality is unimportant because dependent packages did not use them, then a teacher turned up saying that old code he uses in his course no longer works.

Example data in particular is something that is much more important for research, experimentation and teaching, and is almost never used by dependent packages.

Let's please not judge what is important solely based on dependent packages.

[szhorvat]

All data must be under the same license

Can you explain why? Is this a CRAN requirement?

[krlmlr]

This is a CRAN requirement. The solution is to move one dataset to a different package.

I hear you're passionate about giving academics the best support they can get. I see tension with ongoing software development.

[szhorvat]

We can remove rfid, it's the only one with the contagious GPL-3 license, and license the package with CC-BY-SA.

If removing a single dataset makes us compliant, that is a good solution I think.

[szhorvat]

Could this be a good replacement? https://networks.skewed.de/net/sp_colocation I'll need time to check carefully. This newer dataset seems very similar and is CC-licensed.

[szhorvat]

@krlmlr How did you determine that the rfid dataset is under GPL-3? Are you relying on the license file, or did you check this independently?

The original publication is here:

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0073970#s5

There is no mention of GPL, only Creative Commons. It is somewhat unclear if the CC license applies to the publication text, the supplementary data, or both. Note that at one point it was a common misunderstanding among academics that "GPL" and "open source" are the same thing, or perhaps that "making free and open" is accomplished by "GPL licensing". We don't know if the original authors actually intended to impose the restrictions of the GPL. GPL doesn't even make sense for data—it is designed for code, and the restrictions it makes (regarding linking and distribution) are not meaningful for data.

Before removing this, let's verify this:

• Let's check PLOS terms to understand whether the indicated license is meant to apply to the supplementary data.
• Let's contact the paper authors and ask about their intention regarding the data licensing.