I’ve recently been looking at some of the DO (Digital
Object) literature, and DOI/DONA, and I’m still not clear what it
amounts to nowadays. I’ve discussed it a bit internally at W3C and
wanted to share my views here as well.
If it wasn’t for the
IoT and ITU context, I’d say it’ll have a negligeable effect on
the existing Internet architecture using URI/http/DNS/IP. But we do
have this context of IoT chaos in terms of standardization (the
main reason why we’re working on the Web of Things, or WoT, as a more abstract layer), so I
think it’s worth discussing a bit more in our community.
Anyway, this is part of an old story.
I remember one particular aspect that was discussed on some of
lists in 2003 when DOI tried to get a URI scheme (i.e. doi:),
after having rejected the idea of a URN namespace (e.g. urn:doi:)
or a non-IETF-tree scheme (like org:doi:) as too second-class
For the novices, DOI is just another persistent identifier
catalog and syntax, using a “global/local” grammar (e.g.
10.101/something), which should ideally be presented as
urn:doi:10.101/something to fit with our architecture (this
apparently works in some tools, even though urn:doi is not a
registered URN namespace) or even as doi:10.101/something, using a
new URI scheme this time (which is also used sometimes in
interfaces or papers, even though it’s not registered with IANA
either, and doesn’t resolve as such).
In the end, today, all DOIs use the DNS and the http URL scheme
like in http://dx.doi.org/10.101/something to enter their
resolution space (which is main reason why they didn’t get their
doi: scheme in the first place, since they have no independent
resolution “running code” of their own for the schema itself, they
just use http with DNS and URL querying).
But the point is, once the initial http://dx.doi.org proxying is
done, they (DOI/DONA) provide a full resolution system independent
of DNS for their own internal PI syntax, with a commercially based
hierarchical registration system comparable, from a distance, to
ICANN/IANA/DNS, with registries, fees, registrants, etc. I haven’t
looked at their pricing, persistency policies, etc. This is a
service that the scholarly community seems to appreciate a lot,
e.g. to be able to dereference an ISBN number into some resources
about it (e.g. the book itself, a summary, some metadata, a link to
a bookstore, etc.). What’s their competition in this space ? Local
universities/librarian portals with URL querying using ISBN ? purl,
ark maybe ? But without a central root (Global Handle Registry)
like what DONA is providing.
It’s also still unclear to me how their hierarchical name space
is organized, e.g. by countries, by industries, both, flat ? using
what semantics ? whatever it is, better or worse than ICANN’s gTLD,
ccTLD and subdomain policies, one paper was saying that e.g. for
China/Russia/Iran, one main advantage over ICANN is that it’s not
run by a California non-for-profit but by a Swiss non-for-profit –
which BTW has ITU singled out in its Statutes as a partner of
choice. There goes down the drain our usual “multistakeholder
matters/geography doesn’t” argument..
An interesting question is why all these countries, somehow
active in ICANN/IETF are at the same time trying to fragment
the root ? I read somewhere that the same person who is/was
overseeing the ITU/DOI work is also on the ICANN GAC for instance,
so there is communication. Another question worth asking is why is
there a recent RFC asking all IETF
RFCs to also provide DOIs ? As if an ietf.org URL was not stable
The ITU move to endorse DOI is clearly political, and there is
no denying that having the top I* headquartered in the states and
having Trump as a potential head of the same states is worrysome
for lots of folks on the globe.
BTW, there seems to be also a metadata stack of some sort being
used/integrated in the DOI system, called indecs, which I haven’t
looked at, and some questions related to URN syntax not being able
to support their needs for structured identifiers.
So far, I haven’t seen any deployment of a custom protocol of
their own (like http or ftp) to justify doi: as a first class URI
citizen, e.g. something that browsers would implement more readily.
It looks like the Handle part of their
system defines that, or used to (there was an hdl: URI scheme
available at some point). Same thing for the deployment of a “bind”
of their own that OSs would have to implement as well, to connect
user agents to their DOI servers directly, without going through
DNS. For now, DOI looks like an alternate root that doesn’t use the
open DNS software infrastructure (but have to use a URL to access
their resolution space).
Once they do that – deploy software that connect directly to
their main resolver/handle, and I think they will if they reach
enough critical mass, what’s behind their doi: syntax doesn’t use
DNS or IP, it’s just a private identifier binding space run by the
DONA/DOI organization/servers and their registries, with a
promise of uniqueness and persistency (maybe I’m just used to W3C
and ICANN and we’re as opaque for newbies, but I can’t say it’s
very transparent to me in terms of who is controlling what, what
ontologies are used, etc., but then again, they don’t really sell
cashable global names like coconuts.net, but dull series of unique
ID/numbers, which look more like IP numbers that domain names from
the outset, except that they are supposed to be assigned to the
same resources forever).
Of course, today, most (all?) doi: identifiers not only use URLs
to resolve their ID (funny enough, at some point, back a year or
so, doi.org resolution was down because DOI had forgotten to renew
its .org DNS registration), but they also return URLs as their main
typed values in DOI records, since that’s the most easily
resolvable ID today on the Internet.
Overall, I’m not too worried, since being on the same root as
the rest of the planet brings more economic advantages that
anything else nowadays, but let’s not forget the IoT context, with
most of the “Things” already in control of governments (i.e. your
fridge, your electrical plug or bulb, your car, are already all
subjects to gov conformance) so an easy target for gov to impose a
particular network interface vs. the Internet stack.