-
Notifications
You must be signed in to change notification settings - Fork 822
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Proposal to add new property /mobileUrl as sub-property of /url #3134
Comments
I am sympathetic but couple points:
How do you define "mobile"? Eventually most URLs will be mobile-friendly.
Do we go back to them all being URLs then?
Also - I think we should link to a cautionary note in the definition which
is that initially at least publishers should bear in mind that we have 10+
years of markup using /url and that not every consuming application
exploits subproperty hierarchies
…On Fri, 8 Jul 2022, 18:25 Alex Jansen, ***@***.***> wrote:
*Context* - This is a proposal from Google based on our experience
consuming schema.org markup and working with similar data from online
merchants. If it were accepted, it would make it easier for us and others
to differentiate mobile optimized URLs from standard (desktop) URLs.
*Proposal* - Schema.org supports a /url <https://schema.org/url> property
for use on /Thing <https://schema.org/Thing>, used to specify the URL of
an item. Online publishers often create separate mobile-optimized landing
pages, and we therefore propose to add a new */mobileUrl* property as a
sub-property of /url <https://schema.org/url> for use on /Thing
<https://schema.org/Thing> to allow consuming systems to better
understand mobile vs non-mobile versions of web pages representing the same
entity.
—
Reply to this email directly, view it on GitHub
<#3134>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGKVLPXSA3NWF4XYD7TVTBQAZANCNFSM53BUN26Q>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
The idea is to use /url for all cases where there is only a single url, supporting both mobile and non-mobile devices, and only use the new /mobileUrl if there is an explicit second different page optimized for mobile devices. Agreed that automatic detection of super-properties might impact existing systems so I would look forward to community feedback. Especially since there will be more cases where the ability to identify a "main" vs "secondary" value of a property could be very useful, for example to identify a primary image for the /image property on /Product. This could be done through a sub-property of /image, or we could add a new standalone property that is not a sub-property (similar to primaryImageOfPage). The latter option would still have a major effect on consuming systems which will have to be updated to process the new property. It would therefore make sense to at least explore a more generic solution. |
Joost Tweeted about a similar thing recently about posterImage and featuredImage being new properties for Article: https://twitter.com/jdevalk/status/1544614542432428032 What makes a sub-property different to a regular property? How about using a sub-class instead. Url->MobileUrl and then providing an array for url with one of the entries using MobileUrl to indicate it is specifically for mobiles. Or would consumers struggle with url becoming an array and dealing with sub-classes? I guess adding a new property is less likely to break existing implementations. |
If discovery and association are the goals, then why not declare the non-moble site in the Sitemap file? |
Thanks, everyone. Ok, lots of things going on here!
So there's some careful wording to find there. We know that it is common for organizations to spin up a whole new site for mobile, regardless of trends and best practice. So having /mobileUrl makes some sense to identify those. On @Tiggerito's suggestion(s) - using subclasses here is tricky. Firstly, the Schema.org datamodel is a delicate balancing act between the HTML5 Microdata syntax and the conventions from W3C RDF. Our representation of URLs as strings is somewhat unusual in RDF usage, where all nodes in the graph can carry URIs/URLs. I am wary of complicating the situation further by having types whose members are the URLs (rather than the people, places, things, sites, pages etc.) identified by those URLs. Secondly the terminology of "array" (presumably from JSON) doesn't quite work here, since our datamodel is common across JSON-LD, RDFa and Microdata, as well as other data environments like RDF-Star, Property Graphs, SPARQL/Turtle etc. When we use arrays in JSON-LD it is typically an unordered list representing repeated use of a property on some item. This stems from early confusion in JSON about key redeclaration (e.g. see here). @jdevalk also made very reasonable enquiries about sub-properties. To answer @Tiggerito's question: a sub-property is just a property, so it is no different to a regular property. The point about sub-properties is just that they indicate reliable patterns of truth amongst descriptions that use them. For example, whenever something 'X' has a brother 'Y', then 'X' has a sibling 'Y'. So we can capture that by saying that the more specific property, let's call it /hasBrother has a more general "super property", "/hasSibling". The subPropertyOf relationship is just an awkward semi-backwards way of saying "super property". Both relate a specific property to a more general one that it implies (given some pair of entities etc etc). In this way sub-properties are very similar to the subtype hierarchies which are much more prominent at Schema.org. You can infer that something which is a OpinionNewsArticle will also be a NewsArticle (or Article, or CreativeWork, or Thing). As you work up the hierarchy things get less informative, and more general. So one problem for us with sub-properties (why we don't go crazy using them everywhere) is that in most of the concrete markup notations available to us (except RDFa), it is pretty awkward to list multiple properties connecting two specific entities. This puts publishers in situation of deciding whether to use the more informative, specific property (which may also be newer, less widely understood) or the more boring general superproperty. It also puts data consumers in position of doing more processing to infer super-properties when they encounter instead a sub-property. To conclude: I think we should add this to Pending, but with some design constraints:
|
@alex-jansen does this meet your usecase? Could you draft a line for the release notes? Queued for staging |
Thanks @danbri this will work. Agreed with the complexities surrounding sub-properties. |
This feels like a slippery slope to defining infinite numbers of URL types based on context, which seems like an oddly arbitrary direction for the standard to travel in. Should we support Not keen on adding arbitrary appendages to the standard just because folks at Google want a simplistic attribute to sniff for, when a little exploration might lead us to a more graceful solution. If we're trying to describe that a certain type of thing is optimally consumed by a certain media, that's nothing to do with the URL - it's a descriptor of a consumption method, which may have a URL... surely? And if the objective here is to map the relationship between 'desktop' and 'mobile' URLs (for publishers who can be bothered to mark that up correctly), we already have a standard; |
FWIW, I fully agree with @jonoalderson that this is not a good idea. We already have standards for this... I feel we should just tell people that need this, that their website is broken. Not clutter schema.org by adding properties for things that shouldn’t exist. |
“We made a mobile site” is a thing that happened in the world, for better
or worse.
There is a fine line between simple and simplistic but if you look at the
proposed wording I have tried to box the usecase in, since taxonomising
kinds of sites more thoroughly is pretty fraught.
We could of course look at subclassing /Website, but then we hit other
usability issues around “is the url for a page different to the url for a
site where the front page is that page?”.
…On Tue, 9 Aug 2022 at 21:00, Jono Alderson ***@***.***> wrote:
This feels like a slippery slope to defining infinite numbers of URL types
based on *context*, which seems like an oddly arbitrary direction for the
standard to travel in.
Should we support DesktopUrl? What about printUrl? Or bookmarkURL? Or
RssFeedUrl? Or OnMySmartToasterUrl? How many URLs can/should/might a
*thing* have?
Not keen on adding arbitrary appendages to the standard just because folks
at Google want a simplistic attribute to sniff for, when a little
exploration might lead us to a more graceful solution.
If we're trying to describe that a certain type of *thing* is optimally
consumed by a certain media, that's nothing to do with the URL - it's a
descriptor of a consumption method, which may *have* a URL... surely?
And if the objective here is to map the relationship between 'desktop' and
'mobile' URLs (for publishers who can be bothered to mark that up
correctly), we already have a standard; rel=alternate. I've definitely
seen discussions elsewhere here for providing sets of 'alternate' URLs
(e.g., in the context of AMP pages); maybe expanding that with descriptors
of their purpose is a cleaner approach? We could even define media queries,
etc.
—
Reply to this email directly, view it on GitHub
<#3134 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGKHCLHTFGSKV36TDY3VYK2ENANCNFSM53BUN26Q>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Yeah, "mobile sites" as a subclass of I also think that constraining this only to product/offer - purely because that's the bit that the Google Merchant Center folks care about - feels like an even slipper slope than conflating URLs. Products and offers aren't special types of things which have or need different types of URLs from the rest of the web. If we want to solve this, I think we should take the time to do so generally, rather than in a way which creates a Google-centric frankenspec. Appreciate you trying to defuse it, but I'm dubious as to how much impact that'll have on a Pavlovian web. |
Twitter link to surface this to non github folk - https://twitter.com/danbri/status/1557101340094078976?s=21&t=2YfSoX-qmWW7oOE-4pnu7w Arrays don’t really exist in schema.org graph, any more than or ‘itemprop’. How would an application trying to be useful to mobile users select from the different things in the array (or from the repeated property the array encodes).
|
http://pending.webschemas.org/LinkRole feels close to what we'd need? |
I second the remark of @jonoalderson : this opens the door to many other sub properties that will clutter the schema ; where's the limit then. Besides, there are other mechanisms in the web stack to cope with agent-dependent representations, e.g. HTTP redirects based on the User-Agent, responsive CSS, or "rel=alternate" links. |
I am increasingly feeling that the generic edge-annotating pattern we
defined for Role is more or less a failure. That makes me wary of building
more things on Role types, until that analysis/cleanup is done.
There is a general mechanism being spec’d at W3C for attaching arbitrary
information to any and all edges in the graph- i.e rdf-star.
https://w3c.github.io/rdf-star/cg-spec/editors_draft.html
I don’t think it’s quite ready but there is a lot to be said for a
mechanism that allows a common property to be shared even while carrying
additional qualifications of various kinds.
:something :url “https//…/x1234”
{| :mobileyNess :VeryMobileFriendly |} .
(In Turtle-star proposed notation the inner annotation is also asserted)
With this kind of thing *potentially* available soon, whatever we do now
feels like it will inevitably be a pragmatic stepping stone.
…On Tue, 9 Aug 2022 at 21:40, Jono Alderson ***@***.***> wrote:
http://pending.webschemas.org/LinkRole feels close to what we'd need?
—
Reply to this email directly, view it on GitHub
<#3134 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGLUUKHCNE56SUBMT3LVYK63ZANCNFSM53BUN26Q>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Yeah, That said... If we're in a position where we're responding to demand from Google to describe URL roles, I can't help but think that there's some responsibility for 'them' to seek to do that responsibly (i.e., using scalable syntax). I'm sure that if Google's documentation described and provided examples of how to mark up roles, then there'd be reasonable (or no less than otherwise) adoption? Did roles fail, simply because there was no demand? And now, what is this, if not demand? |
On Tue, 9 Aug 2022 at 22:17, Jono Alderson ***@***.***> wrote:
Yeah, Roles are tricky, and adoption is low. 👍
That said... If we're in a position where we're responding to demand from
Google to describe URL roles, I can't help but think that there's some
responsibility for 'them' to seek to do that responsibly (i.e., using
scalable syntax). I'm sure that if Google's documentation described and
provided examples of how to mark up roles, then there'd be reasonable (or
no less than otherwise) adoption?
Did roles fail, simply because there was no demand? And now, what is this,
if not demand?
They look weird to people who understand the RDF / graph perspective, and
they look weird to people who don’t. The pattern with the property name
repeated as it goes into and out of a Role node is pretty unusual and
counterintuitive. Every extra markup structure we introduce creates scope
for new kinds of bad markup. I am not aware of anyone claiming they have
code solidly consuming the Role pattern, in a way that allows it to not
matter whether a normal short factoid is being consumed, or an annotated,
mediated, edge-role-edge structure.
—
…
Reply to this email directly, view it on GitHub
<#3134 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGK2NYUF5J6JBZ3DPT3VYLDFZANCNFSM53BUN26Q>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
What if we created a type This could also be useful for images, to point to other mine types, and potentially more applications. That’d be a lot cleaner and useful in the long run. |
@danbri said:
For some value of soon. Presuming that the proposed RDF-star WG is chartered (voting should be done), then it's ~2 years before a REC is out. Not a specific work item, but there is a JSON-LD-star draft that would likely also come out, but so far no work on RDFa-star much less Microdata-star. Of course, this does enable some useful use-cases such as this, and what was attempted with the Role class, but some more complicated use cases may involve the introduction of more blank nodes as discussed Triples and Occurrences EXAMPLE 8:
|
I have to say that for a big part I agree with what @jonoalderson and @jdevalk have said. Though to be honest, my initial thoughts when I read this was: "Oh no, not another crucial thing site owners/devs/consultants can and most likely will screw up. I'm going to advice my clients not to use this just to prevent any future issues".
I fear adding this could lead to many more issues than it's worth. |
Please!! In the documentation clearly define "separate site". E.g., not just a responsive page, address of it must be a separate domain or sub domain can one have two sites on the same domain? Multi-site Wordpress allows different sites in a folder like hierarchy. But my reading of google indexing documentation for SEO is that they treat all sub domains as new domains but "all sites" may not be on different domains. Some clarity here need to exist in the documentation if this contentious proposal is moved forward. |
I have started a design document with several options for addressing this use case, based on the discussions so far. It is long and a rough draft, and I haven't checked it over for mistakes and repetition but I figure best to share in initial form here: Indicating mobile version of a site's URL I think the points made here about the relationship with HTML level metadata, especially from typed links, are important. However I do believe it is worthwhile being able to map such information into Schema.org, just as we can do with iCal files and other dedicated domain-specific formats. At this point (since the draft is so rough) I would discourage detailed critiques, in case I've mis-stated something or just otherwise goofed up. That said, do please take a look! Schema.org has a lot of hidden baggage that constrains our options, and one lesson from this is that they ought to be represented explicitly somewhere. |
FWIW "design 2" in the draft is I think pretty close to what @jdevalk is asking for here:
However I couldn't make out if alternateURL was a type or a property here. With LinkRole the "alternate" aspect is covered by the linkRelationship, rather than baked-in vocabulary, hopefully leaving room for other cases to be handled without additional schema changes. The doc has an example of this, sketching a rel="canonical" from the mobile site back to the main one: "relatedLinkSpec": [
{
"@type": "LinkRole",
"fromLink": "https://www.example.com/product2213",
"toLink": "https://m.example.com/product2213",
"linkRelationship": "alternate",
"media": "handheld"
},
{
"@type": "LinkRole",
"fromLink": "https://m.example.com/product2213",
"toLink": "https://www.example.com/product2213",
"linkRelationship": "canonical"
}
] See also IANA Link Relation Types registry and this Google blog post it references. |
I was thinking of a property at the time I think. Reading your doc - and mind you I do appreciate all the explanation in it, it's super helpful, so thanks for that!! - I'm afraid that this is all overly complex. To be honest, whichever option we chose will only cause more people to "do it wrong". Given how hard people are finding things like |
@alex-jansen I keep coming back to the fact that having a version of |
I still feel adding web page alternate/media/hreflang meta data to schema.org is a route we shouldn't take (although I do understand the appeal of it). Not only do tons of sites make horrific mistakes with the alternate/media/hreflang/robots/canonical info they publish but on top of that for many it will be close to impossible to get that information into their markup. Reason being that this type of information often isn't part of the info a CMS stores. More often than not information published in the Having said that, if we leave out the alternate/media/hreflang part of things I do feel that with a slight tweak we might be able to get what @alex-jansen is looking for while also accommodating @jdevalk's wish to be able to point to a json-ld representation of a page. What if instead of adding Given all the possible scenarios @danbri described in his document (thanks for that by the way, it really helped) I feel that this solution has the least downsides (mainly at the consumer side) while also keeping it simple enough that it actually might get adopted at a reasonable scale, without tons of sites making a mess of things.
|
Oh, and maybe |
Thanks everyone! Just jumping in on this last suggestion for now: If all we write is {
"@context": "https://schema.org/",
"@type": "Product",
"@id": "#some_offer_1234",
"url":"https://www.example.com/product2213",
"alternateUrl": "https://m.example.com/product2213"
} This does not tell us anything different about the second URL, and code written in the past wouldn't know to look for it. JSON-LD represents repetition of a property with JSON arrays, so we can just repeat several "url" values: {
"@context": "https://schema.org/",
"@type": "Product",
"@id": "#some_offer_1234",
"url": ["https://www.example.com/product2213", "https://m.example.com/product2213"]
} This would make both URLs visible to code based on Schema.org 2011-2021. It would not tell us anything at all about what the difference between them amounted to, though. I do very much hear you (all), and the concern that whatever we do doesn't add an additional layer of complexity and duplication to what we have in HTML. However I do continue to believe that there is value in having a way to write down in Schema.org (RDF, JSON-LD etc.) the information communicated by a collection of HTML pages. Schema.org does not itself say much about "what to say where", e.g. should an e-commerce site with 5 million pages declare their phone number or logo on each of those pages? Any thoughts on RFC 9264? It seems to pull together a kind of data model for the information scattered across pages in rel=xyz syntax. We are also pretty close here to the notion of Schema.org feeds/sitemaps/dumps, in which a single larger Schema.org file might summarize the contents of many different web pages. |
Before diving into the RFC documentation you shared @danbri, I have some questions I think are of relevance in regards to the feeds/sitemaps/dumps... Why the desire for something new to accumulate repetitive info in as opposed to 'simply' digesting cross-page referenced entities? To give a specific example (we all can easily come up with other ones as well):
Now the only reason I tend to have websites publish the same info over and over again is because search engines have very specific requirements (as well as a lot of recommendations) to be eligible for their search result features and without including such repetitive info on each Product page there will be tons of errors/warnings and possibly even a loss of eligibility for the desired search result features. As a publisher I'd be an extremely happy camper if I'd never have to write any repetitive info ever again if instead I could simply refer to any entity, no matter where it 'lives' on a site, e.g.
What I've been wondering ever since the first mention of a sort of 'feed' is why wouldn't the above method be enough (outside of the fact search engines having to be willing to adopt cross-page referencing), why the need for an even more aggregated form? Personal note: |
I don't fully agree with that statement as IMHO Now If that's too much to ask of consumers (which I can imagine) I'm not sure how to move on from here as I feel it's too much to ask of publishers as well due to the financial implications this will have (of course I'm not talking about large corporations here but more SME sized organizations), as well as the amount of technical issues this will most likely generate.
Is this really that much of an issue? (sincere question, no sarcasm intended) The web evolves, technologies change, organizations adapt and move on to the next new thing/issue. IMO a new property isn't something so crucial consumers wouldn't be able to adapt without too much trouble - or at least not compared to the effort the rest of the web has to put in to successfully integrate alternate/media/hreflang meta data into their markup. Again, I do understand the appeal of having such info in markup - heck, in the future I might even find a purpose for it in my own projects as well - though I remain sceptical to create vocabulary for it if the intent is to have this be adopted at scale based on the amount of websites that to this day still have issues implementing alternate/media/hreflang/canonical/robots meta data in the |
To add some context to the technical/logistical/resourcing difficulties around that, even in Yoast SEO (in WooCommerce and Shopify) we'd struggle to help users to describe the differences between (and varying roles of) their alternative URLs. There are huge UI / definition challenges around this, and absolutely zero standardization in how such variations are handled at present. WordPress (rightly?) doesn't even have a concept of different URLs for different intended device consumption. That'll realistically mean no correct/meaningful adoption across the whole of WordPress and Shopify. Sure, other platforms exist. But if this approach is hoping for meaningful adoption, a significant portion of the web isn't designed to be able to handle it. As Jarno points out, hreflang is 'hard' enough, and this is an even more complex version of the same type of problem. I think signposting the existence of an alternate URL is a sensible middle-ground; and let the respective |
Just playing with combining some ideas. How about an alternateUrl property that accepts URL and LinkRole. And some changes to LinkRole:
An example including the different ways links can be added to alternateUrl:
And where LinkRole could be used on its own:
|
This is a great discussion on an obviously complex topic, also illustrated by previous more or less failed attempts. It might be worthwhile to note that at Google we do see a substantial percentage of eCommerce sites with separate mobile-optimized versions, which is why we provide the ability for merchants to supply separate mobile URLs in our Product data specification. Other players such as Meta and Microsoft support the same attribute as well. I hope this at least illustrates that supporting alternate URLs would be a valuable addition to Schema.org. Another option would be using AI or other mechanisms to infer which sites are mobile versions of other sites, but allowing the explicit identification of related URLs in the structured data markup itself seems more predictable and gives more control to site owners. Taking @jonoalderson's comment, this discussion indeed goes beyond just mobile-optimized pages. At Google we for example also support store-specific offer URLs for merchants with physical stores. Taking @jdevalk proposal then to add a new additionalUrl property with the ability to optionally specify its role for those sites that can support it seems eminently reasonable to me and worth iterating on. |
Thanks @alex-jansen! I think the challenge we have with an {
"@context": "https://schema.org/",
"@type": "Thing",
"@id": "#this"
"url": "https://www.example.com/product2213",
"alternateUrl": [
"https://other.example.com/product2213",
{
"@type": "LinkRole",
"url": "https://m.example.com/product2213",
"linkRelationship": "alternate",
"media": "handheld"
},
{
"@type": "LinkRole",
"url": "https://www.example.com/es/product2213",
"linkRelationship": "alternate",
"inLanguage": "es"
},
{
"@type": "LinkRole",
"url": "https://www.example.com/fr/product2213",
"linkRelationship": "alternate",
"inLanguage": "fr"
}
]
} Even better, readibility wise, would be to add two types of link relations, {
"@context": "https://schema.org/",
"@type": "Thing",
"@id": "#this"
"url": "https://www.example.com/product2213",
"alternateUrl": [
"https://other.example.com/product2213",
{
"@type": "LinkRole",
"url": "https://m.example.com/product2213",
"linkRelationship": "alternate-device",
"media": "handheld"
},
{
"@type": "LinkRole",
"url": "https://www.example.com/es/product2213",
"linkRelationship": "alternate-language",
"inLanguage": "es"
},
{
"@type": "LinkRole",
"url": "https://www.example.com/fr/product2213",
"linkRelationship": "alternate-language",
"inLanguage": "fr"
}
]
} |
I've seen/dealt with this on ecommerce websites in the past (same goes for the mobile urls as well as urls created for different merchant platforms) though in many occasions this involves a custom solution either created by the organization itself or a third party (sometimes even including external managing tools and hosting). Many of these solutions often are 'hacked' into an already existing platform and therefore more or less operate outside the regular processes of how a website is generated (marketing tends to disrupt internal development processes) - which admittedly doesn't have to be any issue for single language/country sites at all. The real trouble starts when there's internationalization involved and one has to start expressing these 'exceptions' via alternate/media/hreflang in the And it's exactly these types of exceptions where the issues with expressing link roles happen a lot. Not so much in regards to the urls provided via any product XML feed (as these are custom) but getting them to 'fit' into what's being expressed in the Again, I rarely see this cause any issues for any product feeds nor for the 'exception' pages themselves but it mostly happens everywhere around it (not necessarily an issue for the merchant/advertisement platform though it can be quite detrimental for your organic search results). |
Maybe it's a naive question @alex-jansen but is there any chance you have any insights into what % screws up their alternate implementations? Reason I ask is because I'm aware my opinion is likely biased by my profession as I've mainly worked on websites that make mistakes (I'm not called upon if all goes well) and so it could well be that % isn't as big as my experiences make me believe. |
Thanks everyone for all the discussion here, and for your insights and perspective. Here's what I suggest we do:
If this seems a viable way forward, I will set things moving. I think it is at least clear that this issue is unresolved and should not delay our upcoming release. |
Responding to @jvandriel's question separately:
Speaking from a Google perspective, we do not have numbers to share at this point. We can say it fits the general longstanding pattern with Schema.org, which is that without supporting tooling, documentation and consuming applications, data quality can be pretty rough. When there are incentives and support, and ideally a clear link to business priorities, data is almost always much better. My understanding is that in the "mobile URL" case, what we see fits this pattern. Google's Shopping Feeds infrastructure is associated (historically at least) with a commercial relationship vendors have with Google for which their data is central. And the feed data is generally pretty good, because the data is used and useful and tied to costs and expectations that are carefully monitored. |
@danbri Assuming that you want to hear from those involved in this thread, per your:
So: I think you're right in your conclusions. |
Thank you, and yes - I would like to hear from those who were engaged in
the earlier discussion!
…On Tue, 20 Sep 2022 at 14:04, Joost de Valk ***@***.***> wrote:
@danbri <https://github.com/danbri> Assuming that you want to hear from
those involved in this thread, per your:
If this seems a viable way forward,
So: I think you're right in your conclusions.
—
Reply to this email directly, view it on GitHub
<#3134 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGN4VRODZ32CVNKQ5JDV7GY65ANCNFSM53BUN26Q>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Glossing over the discussion, I am reminded that Google uses positioning for image expected aspect ratios.
So a simple fix might be...
|
My hesitations are solely based on the fact I foresee all kinds of technical obstacles for website owners due to limitations originating from how different content management systems work. As long as we take this into account during further discussions I'm all for continuing the discussion as I do see value in getting this resolved. In the meantime I'll try to reach out to several companies that have specialized crawling software in hopes of having them join future discussions so that we can make sure that whatever it is we come up with (outside what is mentioned in the html) can be checked and verified. |
This issue is being nudged due to inactivity. |
Ok it’s on me to propose something clean that meets the needs discussed
here in 2022.
…On Tue, 10 Jan 2023 at 21:31, github-actions[bot] ***@***.***> wrote:
This issue is being nudged due to inactivity.
—
Reply to this email directly, view it on GitHub
<#3134 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABJSGP36IMDYTWVD4CNKHLWRYLQ3ANCNFSM53BUN26Q>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Context - This is a proposal from Google based on our experience consuming schema.org markup and working with similar data from online merchants. If it were accepted, it would make it easier for us and others to differentiate mobile optimized URLs from standard (desktop) URLs.
Proposal - Schema.org supports a /url property for use on /Thing, used to specify the URL of an item. Online publishers often create separate mobile-optimized landing pages, and we therefore propose to add a new /mobileUrl property as a sub-property of /url for use on /Thing to allow consuming systems to better understand mobile vs non-mobile versions of web pages representing the same entity.
The text was updated successfully, but these errors were encountered: