Re: Schema.org extensions versus core

On Wed, May 6, 2015 at 8:16 PM, Tom Marsh <tmarsh@exchange.microsoft.com>
wrote:

>  In general, I really like the attempt to quantify usage as a way to
> qualify something for the core. I’m not sure the two metrics (# of
> websites, # of weekly users on consuming apps) are optimal for this, though.
>
>
>
> For # of weekly users, I think Martin’s option of “total number of human
> visitors to the web sites that implement the conceptual elements” is a
> better measure. I would expect this to include search engines showing
> results for the conceptual elements and would tweak it from “visitors to
> the web sites…” to “visitors to the web pages…”, but otherwise, I think
> this is a great measure. It is basically a web-wide popularity measure,
> independent of whether a given consumer has yet paid attention to the
> domain.
>
>
>
Let me respectfully disagree. The amount of traffic the publisher of the
data has on their own site is not very relevant. If there are a lot of
people looking for that structured data amongst the people who are
consuming it --- search engines and the like --- then it is important.



>  Wes, do you think the above would give you a way to measure the usage of
> schema:Course?
>
>
>
> For websites, what about cases where there are a few very high-volume
> sites. For example, if there were no eCommerce vocabulary already in the
> core, wouldn’t the existence of Amazon, eBay, and Walmart by themselves be
> sufficient to justify having one? If so, perhaps we can simply eliminate
> this criterion in favor of the (proposed adapted) # of weekly users
> criterion?
>

I would like for us to focus the core on vocabularies that will be used by
a lot of sites. Not just a few.


>
>
> The compactness constraint is a tough one. As worded, I think it may
> encourage people to “trickle in” their changes. I.e., it doesn’t prevent me
> from adding 5 terms per release for 10 releases. It would be nice to say
> something about how large a given “domain” (products, autos, medical, etc.)
> should be allowed to get instead, but then you need to define what a domain
> is and how to tell which terms are within it. Anyone have a good way to do
> this?
>
>
>

Yes, we have. We have not let that happen.


>  *From:* Wes Turner [mailto:wes.turner@gmail.com]
> *Sent:* Wednesday, May 6, 2015 3:04 PM
> *To:* Ramanathan Guha
> *Cc:* public-vocabs@w3.org; Martin Hepp
> *Subject:* Re: Schema.org extensions versus core
>
>
>
> So, for example, schema:Course:
>
> How would one quantitatively justify schema:Course in terms of impressions?
>
> On May 6, 2015 4:44 PM, "Guha" <guha@google.com> wrote:
>
>  I am referring to users of applications that consume the data (like
> search engines).
>
>
>
> guha
>
>
>
> On Wed, May 6, 2015 at 1:50 AM, Martin Hepp <martin.hepp@unibw.de> wrote:
>
> Dear Guha:
> Thanks for this important guideline! One question - could you please
> clarify what you mean with:
>
> > 2. It must have at least 10m weekly users. Preferably 100m
>
> Does this refer to the number of times a type or facet is relevant for a
> search engine query? Or the total number of human visitors to the Web sites
> that implement the conceptual elements?
>
> Martin
>
>
>
>
> > On 06 May 2015, at 03:59, Guha <guha@google.com> wrote:
> >
> > There has been a request to clarify when something should go into an
> extension versus when something should go into the core. Here is a first
> stab at clarifying that.
> >
> > For something to be in the core, the following conditions must be
> satisfied:
> >
> > 1. There must be at least a 1000 sites that will use it. Preferably
> 10,000+
> > 2. It must have at least 10m weekly users. Preferably 100m
> > 3. The vocabulary must be relatively compact. Less than 20 terms.
> >
> > Of course, these are not hard constraints. We also recognize that
> vocabularies evolve and more usage than planned might happen. We expect
> terms or entire vocabularies to move from the extensions to core and vice
> versa.
> >
> > This is a start of the discussion.
> >
> > guha
> >
>
>   -----------------------------------
> martin hepp  http://www.heppnetz.de
> mhepp@computer.org          @mfhepp
>
>
>
>
>
>