New Vocabulary for Datasets

[dustmop]

There's unavoidable confusion whenever we talk about data (what is "data", what does the word mean, how is it distinct from similar terms) but even still, I believe we can improve some of the terms used in Qri to help tell better stories. One particular example involves Dataset, which is distinct from DatasetPod, which itself has a field called Data. DatasetPod is part of the recent move to naming things "Pod" when they are Plain-Old-Data, using all basic types, that can be trivially serialized. Yet, it and Dataset are still distinct from the field Data, and in fact it's possible to have either Dataset or DatasetPod without any Data, which is strange.

Furthermore, there are other terms like Meta and Structure that both talk about Datasets, but in different ways. It's important to have good explanations for what each of these terms mean relative to one another.

My suggestion is to strongly embrace analogies to html, a well-understood and often talked about technology. The main change is to rename the field Data to Body, matching how an html document may have meta tags, but that the body is the main thing a consumer cares about reading. This leads to the following comparisons in Qri:

• Structure :: XSD (xml schema definition, describes an xml / html file's structure)
• Meta :: meta tags (contains descriptions, keywords, search engine stuff)
• Body :: body tag (the document's real contents)
• Dataset :: webpage

Most importantly, this last point reveals we can talk about Datasets that don't have bodies, perhaps because they haven't been loaded yet, or because we only have a reference to them. This is similar to the HTTP method "head", which retrieves information about a webpage without getting the full body. No more talking nonsensically about "Datasets without Data"; saying instead "Datasets without Bodies" makes much more sense.

[dustmop]

Some implications:

qri add --data=stuff.json

should become:

qri add --body=stuff.json

And

qri data me/test_name

should become:

qri body me/test_name

or

qri getbody me/test_name

maybe even

qri get me/test_name

which again resembles HTTP.