Re: SHOULD/MUST fragment identifier definition question

On Mon, Dec 10, 2012 at 8:10 AM, Henry Story <henry.story@bblfish.net>wrote:

> A question to supporters of SHOULD/MUST:
>
> This is to be found in the terminology section:
>
> for MUST:
>
> http://dvcs.w3.org/hg/WebID/raw-file/d21603d3972a/spec/identity-respec.html#terminology
> [[
> A WebID is a URI with an http or https scheme, which contains a  URI
> fragment identifier (i.e. a #id ) and which uniquely denotes an Agent
> (Person, Organization, Group, Device, etc.). The URI without the fragment
> identifier denotes the WebID Profile page.
> ]]
> should the text contain a MUST there, or is the above strong enough?
>
> for SHOULD:
>
> http://dvcs.w3.org/hg/WebID/raw-file/http-hash-uri-should/spec/identity-respec.html#terminology
> [[
> A WebID is a URI with an HTTP or HTTPS scheme which uniquely denotes an
> Agent (Person, Organization, Group, Device, etc.). This URI SHOULD include
> a fragment identifier (a string after a "#" character).
> ]]
>
> Does a hash URI require a string after the hash character?
>
> Facebook for example does not have such a string as you can see here:
>
>   curl -H "Accept: text/turtle" http://graph.facebook.com/bblfish
>
> Also is the terminology section the normative one?
>

Yes, it is.

The ABNF for URI [1] does not prevent the fragment from being empty, so
empty frag hash URIs are legal. Note however that in javascript in the
browser you can't use window.location.hash to know whether it's a hash URI
or not because it returns the empty string whether you have a hash with the
empty frag ID or or no hash at all (you'd have to parse windown.location
manually, but I'm not sure if people would think to do that). Also, I would
not be surprised if some APIs generating URIs out there would not append a
hash unless a frag string is passed in, but that doesn't concern us and
doesn't mean we have to disallow the empty frag. It does look odd and
people might be tempted to remove it thinking it's useless. Having an
explicit fragid makes it obvious that it is intended and it's not a
mistake, but that's just usability, nothing that needs to be required in
our spec. I'm of two minds about this: Henry, what change do you propose
exactly? what would you remove from the definition? we could say "A WebID
is a URI which contains a fragment identifier or ends with '#'".

Steph.
PS:

[1] http://tools.ietf.org/html/rfc3986#appendix-A