Proposing to remove:
Note existing discussion in #122
I don't agree, but I might be alone. I find that all of these changes which are just (if I understand how the versioning system works) patch versions to be pretty big changes (some even backwards incompatible like renaming stuff:
At least knowing what version a data package is, is very helpful for those who build data package parsers. My vote would be to go the other way and make this a MUST (but I'm also willing to accept that I might be the only one who is dumb enough to rely on versioning of often times hand-made descriptor files).
To be clear: I agree this could be useful. The issue is will anyone observe it? I haven't added it for a single one of the data packages i've created. Sure tooling could add it but none does afaik and you'd have to worry about tool having wrong version. Finally, would anything consuming actually know or care (at least atm).
So my point is not that this couldn't be useful but that it is unlikely to, at least for quite some time and I'd prefer not to have stuff in the spec that we don't actually observe - this is a general principle of spec stuff that "its easy to want things and hard to have them"
I'm going to implement this - i.e. remove the field. It is not required, it is only vague used, i think it is probably unreliable so little us to consumers etc. My guess is that if you really care you'll probably do some datapackage.json inspection to try and guess what is happening.
[dp][s]: remove `datapackage_version` - fixes #140.