From: worley@ariadne.com (Dale R. Worley)
To: "Soni \"It/Its\" L." <fakedme+ietf@gmail.com>
In-Reply-To: <99f17eb1-686c-4a57-b4e6-531517d2b9cb@gmail.com>
 (fakedme+ietf@gmail.com)
Sender: worley@ariadne.com (Dale R. Worley)
Date: Thu, 22 Aug 2024 12:05:45 -0400
Message-ID: <87ttfcwnja.fsf@hobgoblin.ariadne.com>
Message-ID-Hash: BSLVEIMP34LIM7RTZT3ZI6RE32ZOS7KT
CC: art@ietf.org
Precedence: list
Subject: =?utf-8?q?=5Bart=5D_Re=3A_=5BWitarea=5D_use-cases_for_meta-URIs?=
Archived-At: 
 <https://mailarchive.ietf.org/arch/msg/art/yWYwrZrgep54AxM4FKFyAaDGb6s>

"Soni \"It/Its\" L." <fakedme+ietf@gmail.com> writes:
> indeed, they have not registered their media types. (as is somewhat 
> common with these things...)

That makes it difficult to argue that fragment identifiers are being
used contrary to their definition.

> however: the official mercurial project uses the fragment identifier on 
> a mercurial repository to specify a "revision" as defined in the 
> mercurial docs (one of a hex string, branch name, tag name, revision 
> number, or some other stuff), while archlinux PKGBUILD uses the fragment 
> identifier with a k-v pair, with the key being one of "tag" (a tag 
> name), "branch" (a branch name), or "revision" (presumably it only 
> accepts a hex string for this one), and finally python expects a bare 
> "commit hash" (hex string).
>
> that is, for the same repo, mercurial expects to see #foo while 
> archlinux expects to see (e.g.) #branch=foo and python expects to see 
> #[the actual hash]

Yes, but ...  Validity comes not from what the repository is, using "the
sem repo" doesn't matter, because a repository can be accessed by many
URIs.  And indeed, it's possible for a media type to specify that many
different fragment identifiers designate the same sub-resource
"fragment".  Validity is about when a URI specifies a fragment
identifier, fetching the resource specified in the URI also gives a
media type, and then whether the fragment identifier is being
interpreted by the fetcher in accordance with the definition of the
media type.

Can it give a specific example of such an incorrect use of fragment
identifiers?  That is, a URI-with-fragment-identifier that is used in
the wild, where if *I* fetch the resource using the URI, and look at the
media type, I can see that the fragment identifier is not being used in
conformance with the specification of the media type?

Dale