Michael,

I sense confusion between DOIs and other forms of stable identifiers.

There are various architectures for stable identifiers -- OCLC's PURL, 
the CNRI Handle System, etc.  There are a number of implementations of 
the Handle System, the most well-known of which is CrossRef's DOI.  As 
you note, CrossRef has a particular scope for what sort of content can 
get a DOI, and those who register DOIs are required to do certain things 
required by CrossRef.

For use case (1), I wonder if the funder really requires only a stable 
URL, not a DOI.  If a funder really requires a DOI for a website that's 
not a journal article, they're asking for something that's not really 
allowed.

For use case (2), I suppose you could run an implementation of the 
Handle System outside of DSpace that allows you to create URLs for the 
bitstreams.  I haven't worked with DSpace and have heard that it's 
difficult to modify; still, I find it surprising if you're the first to 
try this.  Perhaps ask around in the DSpace community?

Use case (3) is the most appropriate, but you'll find that CrossRef's 
membership fees are quite high.  You might set up a local implementation 
of the Handle System to create your own handles rather than use 
CrossRef's.  That way you get the permanent identifier without the 
fringe benefits of CrossRef membership, its obligations, and its fees.

Hope that helps,

Kevin

Michael B. Klein wrote:
> Hi all,
> 
> I've been investigating the possibility of assigning DOI names to various
> resources. We have three different use cases, and given the structure of the
> DOI Registration Agency system, I'm not sure what the best way is to
> proceed.
> 
> The use cases:
> 
>    1. We've had several inquiries from faculty whose research funding
>    requires them to publish a web site, and identify the site with a DOI name
>    for citation purposes.
>    2. We'd like to assign unique DOI names to specific bitstreams within our
>    institutional repository. Despite the fact that DSpace uses a handle server
>    internally, the handles it assigns resolve to metadata/landing pages, and
>    there doesn't seem to be a good way to create a reliable, persistent link to
>    a full text PDF that will migrate easily to a different IR system if and
>    when we choose to move away from DSpace.
>    3. We're investigating the possibility of publishing a couple journals,
>    and would want to use DOI names to identify articles and related content.
> 
> However, it looks like each existing DOI Registration Agency has a specific
> subset of content and services they work with -- journal articles for one,
> datasets for another, etc. -- and I'm not sure how to go about finding an
> agency that will let us assign suffixes in a way that works with our varied
> content.
> 
> Any suggestions/experience would be greatly appreciated.
> 
> Michael