Comment by vivzkestrel

since we are on the topic and since your article clearly mentions "Normalize your data unless you have a good reason not to" I had to ask. I am trying to build a news aggregator and I have many websites. Each of them has slightly different format. Even though I use feedparser in python, it still doesn't change how some of them put html text inside content and some of them break it down into a separate media xml attribute while retaining only basic textual summary inside a summary attribute. Do you think it makes more sense to store a single rss item as a column inside postgres or should it be stored after parsing it? I can see upsides and downsides to both approaches. Store it as XML and you have the ability to change your processing logic down the line for each stored item but you lose the flexibility of querying metadata and you also have to parse it on the fly every single time. Store it in multiple columns after processing it and it may require different logic for different websites + changing your overall python processing logic requires a lot of thinking on how it might affect some source. What do you guys recommend?