If the persons who compiled a data set (or made a latex package or or or) published a paper on that thing, by all means, cite it.

If they didn't, but they do have a web page on the thing, then give them credits in the currency of the web: links.

GP has a point: if the creators want scientific credit, they can easily get that via ArXiv (or, of course, a peer-reviewed publication). If not: increase their PageRank.