I think you’re both right, and also both are missing the point.
Using LLMs to parse whois data is okay in the meantime (preferably as a last resort!), but structuring the data properly in the first place (i.e. RDAP) is the better solution in the long run.
