W3C home > Mailing lists > Public > public-csv-wg@w3.org > April 2015

Re: [csvw] Remove requirement to merge metadata (#511)

From: Ed Summers <ehs@pobox.com>
Date: Sun, 26 Apr 2015 07:51:46 -0400
Cc: David Booth <david@dbooth.org>, Rufus Pollock <rufus.pollock@okfn.org>
Message-Id: <B4D9F2E7-B8CB-4E9E-ACEC-77231840CBC2@pobox.com>
To: "public-csv-wg@w3.org" <public-csv-wg@w3.org>
+1 to remove metadata merging from me too. I think it would greatly simplify the construction of clients, and it will be substantially less rope for data publishers. I really like David’s point about publishers doing the merge once (by whatever means they desire) rather than requiring all clients to do it.


> On Apr 22, 2015, at 2:41 PM, Rufus Pollock <rufus.pollock@okfn.org> wrote:
> ++1 from me on removing merge from the spec. I've always been a -1 on it (though an insufficiently vociferous one given my role as editor).
> Rufus
> On 22 April 2015 at 20:37, David Booth <david@dbooth.org> wrote:
> +1 to NOT merge metadata.
> Merging really is not needed.  Merging adds complexity not only to implementations but also to the user's mental model of how it works.  It makes more sense for the metadata publisher to do any necessary metadata merge ONCE, prior to publishing a new (merged) metadata file, rather than forcing all users of that metadata do merge it dynamically.
> Thanks,
> David Booth
> On 04/22/2015 02:07 PM, Ivan Herman wrote:
> FYI for those in the group who may not look regularly at the github
> issues but may consider an implementation. Thus may reduce the
> complexity of the implementation significantly.
> Ivan
> ---
> Ivan Herman
> Tel:+31 641044153
> http://www.ivan-herman.net
> (Written on mobile, sorry for brevity and misspellings...)
> Begin forwarded message:
> *From:* Gregg Kellogg <notifications@github.com
> <mailto:notifications@github.com>>
> *Date:* 22 April 2015 19:46:55 CEST
> *To:* w3c/csvw <csvw@noreply.github.com <mailto:csvw@noreply.github.com>>
> *Subject:* *[csvw] Remove requirement to merge metadata (#511)*
> *Reply-To:* w3c/csvw
> <reply+0007f21366923576385b67e55029666f10b7cfa2f70886cd92cf00000001114fa08f92a16
> <mailto:reply+0007f21366923576385b67e55029666f10b7cfa2f70886cd92cf00000001114fa08f92a16>>
> During the course of simplifying things subsequent to our London F2F,
> the role of metadata merging has narrowed to the point where it might
> make sense to remove it altogether.
> Originally, the idea was that metadata for CSVs might be pulled from
> several different locations (user-specified, per-file, and
> per-directory), but issues #448
> <https://github.com/w3c/csvw/issues/448>, #314
> <https://github.com/w3c/csvw/issues/314>, #293
> <https://github.com/w3c/csvw/issues/293>, and #200
> <https://github.com/w3c/csvw/issues/200>).
> Currently, the process involves either starting with metadata, in
> which case what is merged is the /embedded metadata/ along with
> user-specified options which typically just affect the /Dialect
> Description/, or start with a CSV, in which case the first Metadata
> file found relating to that CSV is used as if it were the specified
> metadata. With the normative rules for embedded metadata, the
> /embedded metadata/ ends up just being column title information, which
> needs to be specified either through |name| or |titles| in the
> metadata file in order to successfully use the CSV. This could be
> replaced by a mechanism which simply uses the found or specified
> metadata without merging /embedded metadata/ and simply ensuring that
> the CSV columns are compatible with the metadata columns using
> language from the existing merge rules.
> Removing metadata merging is a substantial technical and conceptual
> simplification, and may simplify the steps required to create a
> conformant implementation.
> This change would affect the following:
>   * minor changes to /Locating Metadata/ in the model document.
>       o /Overriding Metadata/ changes to /Overriding Dialect Description/
>       o /Embedded Metadata/ is removed and the discussing of embedded
>         metadata is moved to say that such a definition would need to
>         define how embedded metadata interacts with found/supplied
>         metadata
>   * Minor changes to /Processing Tables/ in the model document.
>       o Simplification of /Creating Annotated Tables/
>   * Remove per-property discussion of merging metadata in the the
>     metadata document
>   * Separate /Normalization/ from _Merging Metadata in the the
>     metadata document
>       o Describe /Normalization/ as a required step when processing a
>         metadata file as part /Creating Annotated Tables/ in the model
>         document.
> I'll follow up with a description of how existing test results would
> be changed as a result of making such a change.
> —
> Reply to this email directly or view it on GitHub
> <https://github.com/w3c/csvw/issues/511>.
> --
> Rufus Pollock
> Founder and President  |  skype: rufuspollock  |  @rufuspollock
> Open Knowledge - see how data can change the world
> http://okfn.org/  |  @okfn  |  Open Knowledge on Facebook  |  Blog

Received on Sunday, 26 April 2015 11:52:12 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:27:47 UTC