Modern sociolinguistics does not have much in common with Boas-Sapir linguistics. The main commonality relates to the subject pool, roughly "various societies in the world". Any population (however defined) can be further subdivided, hence "American society" can be divided into age-based or geography-based subsets. The Boas-Sapir tradition "focused" on a particular subset, those societies that are least well-known and most-different compared to American and European society, but more specifically on unwritten and endangered languages.
In Boas-Sapir tradition, the linguist would engage in a long-term study of the structure of some such society, and in the course of the investigation would gather data for a trilogy of grammar, dictionary and texts. The subject matter of the texts was generally whatever constituted "traditional culture". The details of data-gathering are not well documented, but it appears that the data generally comes from intense one-on-one collaboration between the linguist and the speaker (not from large 10-minute surveys of randomly selected individuals).
To understand the history of modern socio-linguistics, you would want to study the history of sociology, then see how those methods were applied to language in the 60's by people like Sledd, Labov and Weinreich. There is still a contemporary practice of language-description that focuses on under-described languages, where a linguist studies a language for years, producing grammars, dictionaries and texts. This is the contemporary area of linguistics that best preserves the Boas-Sapir tradition.
