Restrict document ID values

[markwalkom]

It seems that there are some chars that are not ideal in document IDs, things such as /, probably others.
This is being discussed on a different level in #9059 and I thought I'd also raise this specific one.

From this forum post.

[jpountz]

+1

[clintongormley]

Actually, I think doc ID values can be pretty much unlimited. As pointed out in the forum post, they just need to be properly encoded in the URL, which all the official clients handle correctly.

[jpountz]

I don't think encoding is enough. For instance we allow documents to have _search as an id but then you can't GET them.

[clintongormley]

You can always fall back to using mget in these cases.

Related to #9059

[dr0i]

Actually, I think doc ID values can be pretty much unlimited. As pointed out in the forum post, they just need to be properly encoded in the URL

Right! If it happen that a value of an _id field is already percent encoded (e.g. a%2Fb) the lookup-query would be .../a%252Fb).
Doing an ID lookup, maybe it would be save to say:
"always percent-encode the ID in an ID-query (take the value of _id , percent-encode it and query that) ".

[dnhatn]

Any document id will be handled correctly with a proper url encoding.