Change Identity schema identifier to unstructured string from URI

[kstreeter]

After working with the current definition of Identity it has been decided that the current definition of Identity.@id is too restrictive due to the need to ensure that all characters in a contributed identity are URI safe. The preference is to define this field in a less structured way, and let implementations manage the details of the representation.

Since @id is defined as a URI, the suggestion is to change this field from @id to xdm:id and designate it as a string versus a URI. This is consistent to how we have represented unstructured identifiers in other places in XDM.

What are the schemas that are affected by the issue

Identity, EndUserIds, Profile, ExperienceEvent (and their extensions)

What are examples of products that are impacted by the issue

Analytics, Campaign, Ad Cloud, Target

[jbeckert]

Is this explicitly only for Identity? Meaning all other usages of @id remain unchanged with the same "type": "string" and "format": "uri"?

[jbeckert]

Does xdm:id for Identity need a "pattern" property to reject ids with prohibited characters, e.g. characters that would mess up the usage of the value in URL path components?

[lrosenthol]

Please don't use xdm:id - it will cause too much confusion with @id.

[fmeschbe]

@jbeckert:

Is this explicitly only for Identity?

Yes, all others keep the @id property

Does xdm:id for Identity need a "pattern" property to reject ids with prohibited characters

That is a good call-out. We should indeed to this at least in the description if not formalizing with a pattern. Your intent would be for this to be the "URL-safe" characters, correct ?

For example we could amend the description to state, that only characters should be chosen for the xdm:id which do not need to be percent-encoded when used in an URL path.

Or we could (in addition) declare this explicit pattern:

pattern = "^[a-zA-Z0-9._~!$&'()*+,;=:@-]+$"

This is the full ability of RFC 3986 pchar except percent-encoded which would be forbidden:

pchar         = unreserved / pct-encoded / sub-delims / ":" / "@"
unreserved    = ALPHA / DIGIT / "-" / "." / "_" / "~"
sub-delims    = "!" / "$" / "&" / "'" / "(" / ")" / "*" / "+" / "," / ";" / "="

I think it is debatable to support all those characters and we might want to actually limit to just the unreserved set.

[fmeschbe]

@lrosenthol I see your point. Unfortunately we cannot change this at this point.

[lrosenthol]

You actually need both - since @id is required in all cases. So in this case, you'd have both @id (which si the URI for the schema) plus xdm:id which is the identity. Which makes this even worse.

I will not be accepting any PR that contains xdm:id.

[fmeschbe]

Created #432 to deal with the character set limitation issue and to thus unblock this issue for release.

[fmeschbe]

Merged #420 and thus closing this issue for now.

[lrosenthol]

@fmeschbe Please UNMERGE #420 ASAP. We are NOT in agreement to accept it.

[fmeschbe]

@lrosenthol working on it.

[fmeschbe]

After some discussion, I am stating:

• We want to cut another internal drop v0.9.3 on July 16, the latest
• We cannot get consensus in this timeframe
• Teams/developers are waiting for the v0.9.3 drop
• The Identity schema is marked stabilizing, so changes have to be expected
• More discussion is needed around the need or not for the @id property and how we want to define it

Gathering this, the plan is:

• Close this issue
• Create a follow-up issue Clarify @id for Identity #434 to track the @id property for Identity schema discussion
• Cut the v0.9.3 drop release as planned
• Schedule the resolution for Clarify @id for Identity #434 for v0.9.4 (next drop release)