Re: Telcon 7th Jan 2015

> On Jan 7, 2015, at 10:58 AM, Ed Summers <ehs@pobox.com> wrote:
> 
> 
>> On Jan 7, 2015, at 3:13 AM, Ivan Herman <ivan@w3.org> wrote:
>> 
>> Personally, I would like to talk about the metadata files, their merge, priority, etc. At the moment, this seems to be the most complicated but also pretty crucial issue.
> 
> Can someone point me at the use case(s) for why this merge/import functionality [1] is required? I think it introduces a fair bit of complexity for someone implementing a client (me). Perhaps that says more about me however :-)

Hi Ed, this was a main discussion topic on today's call, and the future of an explicit "import" property is in question, as it does indeed create some complexities. It came about from discussions at the Santa Clara F2F; see discussion leading up to the resolution [3].

The motivation, as I recall, was to prevent doing HTTP requests for too many locations, and follow the DRY principle. In this case, there may be file-specific metadata which "imports" more generic metadata, so that common annotations, such as publisher, and so forth, can be made in a common location. Otherwise, the metadata location algorithm could be interpreted to look at User-specified, Embedded, Link-header, file-specific, and directory-specific locations (later three I call "found"), requiring a total of four different HTTP gets, where only one or two may be necessary. This is also discussed in issue #145 at some length [4].

Ivan points out the potential to blow up an implementation, due to unknown recursive calls for imported metadata files, particularly in an asynchronous environment. I think we agreed to settle the handling of user-specified, embedded and found metadata and then consider eliminating the import property, or at least evaluate it's importance.

> For some context I’m doing some work with the HathiTrust Research Center [2] making the case that they should share their worksets of book data using CSVW, since they are already making data available as CSV and are considering making it Linked Data friendly.

Great that you're doing a client! I'm doing some work on test cases, but that will likely take some time to emerge.

> //Ed

Thanks for the feedback, and perspectives such as yours are very valuable to the WG process.

Gregg

> [1] https://w3c.github.io/csvw/syntax/index.html#locating-metadata
> [2] http://worksets.htrc.illinois.edu/worksets/
[3] http://www.w3.org/2014/10/27-csvw-minutes.html#item09
[4] https://github.com/w3c/csvw/issues/145#issuecomment-68766764

Received on Wednesday, 7 January 2015 21:11:34 UTC