Hi Melvin,
(not on the TAG, just adding information…)
> On 02 Jul 2015, at 21:40 , Melvin Carvalho <melvincarvalho@gmail.com> wrote:
<skip>
>
> It also ocured to me that we ( working with the team at DIG / MIT ) have been using rel="meta" for quite a while now (with good results), and I wonder if that will become standardized in Linked Data Platform v.next, and is perhaps relevant here. This stuff isnt in Linked Data Platform V1 but may be in V2 (there's been about a dozen people interested in carrying on the work).
>
> For our work, link headers are always used where possible. But some places (e.g. github) dont give you that kind of access.
>
> Now, looking at csvw, I wonder should we be using rel="describedBy" instead? The principle behind it is adding meta data to Linked Data Platform files, much in the same way UNIX adds inode meta data.
>
> In our implementations of meta data we do have standard names such as ,meta (rather than, metadata.json). We also are thinking about rel="acl" for access control. Tho acl was originally tied together with the acl.
>
> I recognize a lot of this, is a new frontier, in many ways, it's not clear cut, and many people wont be in a position to have a view at all. But it is really helpful to hear how others are doing this, and hear ideas, to maybe be able to get our ducks in a row! :)
Actually, this is already part of the spec.
The current approach (see, eg, [1] for the details) is a staged approach with:
1. metadata supplied by the user of the implementation that is processing the tabular data
2. metadata in a document linked to using a Link header associated with the tabular data file
3. metadata located through a site-wide location configuration
(there is an option for a fourth, non-standard approach of including metadata in the CSV file itself, that is a bit special.) Each step is 'overriding', ie, if metadata is found for a specific CSV file through, say, step #2, then step #3 is not even considered.
The current discussion is whether or not using .well-known for step #3. For a data publisher that can set a rel="describedBy" in the HTTP response header (which I think everybody agrees is the best option) step #3 is actually irrelevant…
Cheers
Ivan
[1] http://w3c.github.io/csvw/syntax/index.html#locating-metadata
----
Ivan Herman, W3C
Digital Publishing Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
ORCID ID: http://orcid.org/0000-0003-0782-2704