RE: [Uri-review] Fwd: Re: Informational RFC to be: draft-paskin-doi-uri-03.txt (updated from -02.txt)

["Larry Masinter" <[e-mail]>]

Some background:

When the work on the "Common Name Resolution Protocol" was
begun, there was serious concern about the IETF endorsing
a mechanism that would wind up with "yet another DNS": yet
another system with a (controversial) for-fee mechanism
which charged an annual fee for assignment of names.
Quite a bit of work went into the development of RFC 2972
(which described the likely business models for naming
authorities and resolution services) before the protocol
(RFC 3367) and the URI scheme (RFC 3368) could be published.
That experience suggests that there is some sensitivity
around these issues, and that clear disclosure of policies
and processes is expected.

Secondly, with regard to 'tel' and 'fax' as precedents:
RFC 2806 maps the URI to a well-defined standard (E.164),
and explains the relationship between the E.164 strings and
local dial strings; how to dial a telephone number is
widely understood. The 'doi' document, which cites 'tel'
and 'fax' as a precedent, doesn't meet their level of
specification.

> Is it now to be suggested that new candidate URI schemes must 
> be tightly bound to an Internet protocol?

I think the candidate URI schemes must pass judgment as to
whether they're well-enough defined. I think the definition needs
to be in, or referenced by, the document that registers it.
There might be some legacy cases where the definition is weak,
but we should fix those. For some URIs, there's no resolution
service at all (such as with cid or mid or uuid) but only
a promise of uniqueness. But if you're going to promise
resolution, then the resolution needs to be defined.

In the case of DOI, there seems to be some promise
of a resolution services ("appropriate") without any
definition of what it is, who might run it, etc.

> We also note that the functional requirements for URNs (RFC 1737) do not
> coincide with those of DOI, e.g. URN encoding. Furthermore, URN syntax
> places additional restrictive syntactic constraints on "doi" URIs.

I didn't see anything in the Internet Draft
about functional requirements for DOIs that would lead one
to conclude that the URN syntactic constraints are
unacceptable. 

     RFC 2396 states that a "URN differs from a URL in that it's [sic] 
     primary purpose is persistent labeling of a resource with an 
     identifier". A DOI on the other hand has a dual purpose: both to 
     provide for the persistent identification of an entity of 
     significance to the intellectual property communities, as well as 
     to enable access to a set of network services.

But every URN scheme has a purpose beyond the purpose of "urn"
in general--else why a new scheme at all? The argument is
illogical.


Section 3 of draft-paskin-doi-uri-03.txt seems to repeat
much of RFC 2396 BNF for no clear reason. The DOI syntax
seems to use "/" for a non-hierarchical purpose, which seems
to counter the general recommendations for URI syntax.

> #4 - Google. It should be pointed out that Google is primarily a harvester
> of documents referenced by "http" URI. Further, and more importantly only
> open sites are typically crawlable by robots. Subscription sites offering
> high value content are generally not indexable through search engines like
> Google.

> To support the contention that DOI is a well-established identifier in an
> active and deployed system the I-D referenced the DOI website for further
> information. As a simple indication of current usage, however, you might
> like to review last month's newsletter from CrossRef - a leading DOI
> Registration Agent:
> 
> 	http://www.crossref.org/01company/10newsletter.html#eodiu
> 
> 	"Total Records in System: 7,458,490 (6,644,724 in January)"

I wasn't taking Google as a sole authority; I was looking for
an independent ("without a beneficiary relationship") assessment
of the deployment of DOIs.

> #6 - Patent/Trademark. First it should be pointed out that there is no
> patent issued for DOI. Second, the intention in registering a trademark
was
> purely to secure the integrity of the DOI system. This matter, however,
> would be best addressed by the IDF.

I think the word 'integrity' can be taken several different ways.
The integrity of a fee-collecting organization can only be secured
if it continues to have an exclusive right to collect fees. And
I don't really want to judge the intent, just the operational state.

While RFC 2026 section 10 only requires identification of intellectual
property rights in standards-track documents, it would seem to me
to be part of the judgment even for an 'informational' document
that registers a URI scheme to disclose something that is not otherwise
well-known.

From what I can see in the newsgroup discussion in
the library community, the use of DOI as a permanent
document identifier is considered, at least by some,
to be controversial, exactly because of the issues
of control, authority, collection of fees, and so forth.

Given the discussion in
http://www.crossref.org/01company/10newsletter.html#gi
of invoices, non-linking fees, etc., I think additional
disclosure is warranted, and that publication without
disclosure would be a bad precedent for the IETF.

Larry






_______________________________________________
Uri-review mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/uri-review