[urn] Re: Registration for the TEI: URN identifier

Further thoughts on TEI.  This message rambles, so don't take any part
of it as my definitive opinion.

I was chair of the working group that defined the "alert" URNs to be
used to carry "distinctive ringing" instructions to/from telephones.  The
definition allowed telcos and equipment manufacturers to create custom
alert URNs that incorporate their domain names.

As the specification was approaching completion, Alfred Hoenes (IIRC)
raised the objection that domain names do not have permanent owners.
This was immediately before an IETF meeting, so at the meeting I checked
with the ADs and they agreed that this was a significant problem.  After
looking into various solutions, we eventually defined a special registry
in which enterprises could register their names for constructing custom
alert URNs.
https://www.iana.org/assignments/alert-urns/alert-urns.xhtml#providers
Currently it contains only the entry "example".

So in that case we did "create a new global company registry".

In the case of TEI, I think we can move forward without creating a new
registry using this argument:  TEIs are going to have a considerably
different usage pattern that alert URNs.

In particular, the knowledge of alert URNs is built into telephones and
telephone networks and those systems are likely to remain unchanged for
years or even decades.  So it is important that the permanence of the
meaning of alert URNs is enforce quite strictly.  On the other hand,
looking at
https://github.com/CycloneDX/transparency-exchange-api/tree/main/discovery
it is clear that the usefulness of a TEI URN is tightly coupled to
accessing the HTTPS server at the <domain-name> in the URN.

The consequence of that is that if the ownership of a domain name
changes and the new owner does not intend to continue to support the
existing set of TEI URNs by maintaining the HTTPS server (and the servers
that its /.well-known/tea document points to), the existing TEI URNs
become meaningless *in practice*, although in theory their meanings are
fixed for all time.

It seems to me that this is an adequate response to the problem of "What
if the ownership of a domain name changes?"  We would likely want to put
some information about resolution in the registration to justify not
worrying about this.

However, looking this over, it seems to me that in reality, TEIs aren't
so much *names* as *locators*; nobody really cares about the TEI for an
object if the corresponding servers aren't accessible (at least in the
future).  That suggests that TEI be promoted from a URN namespace to its
own scheme and stated to be a URL.  So instead of

    urn:tei:<type>:<domain-name>:<unique-identifier>

write

    tei://<domain-name>/<type>/<unique-identifier>

That would eliminate all of the philosophical problems.  The scheme
"tei" hasn't been registered.  The procesdures for registering a scheme
name are in sec. 7.2 of RFC 7595 and don't look much more onerous than
registering a URN namespace.

Dale

_______________________________________________
urn mailing list -- [email protected]
To unsubscribe send an email to [email protected]