#microformats 2015-12-01

2015-12-01 UTC
#
Calli
OK. Need to think about it a bit. Did you consider a broader restriction:
#
Calli
Presence of any explicitly marked up p- property on the item (not element) prevents implied p- properties Presence of any explicitly marked up u- property on the item (not element) prevents implied u- properties
#
kylewm
tantek: could you clarify what is the use case here? "Due to indiewebcamp.com/edit use-case, this now makes sense for all implied properties."
#
tantek
kylewm - too long ago to easily reload that context in my head now / asynchronously - can you re-raise in person on Thursday and we can discuss then?
#
tantek
Calli - I did, and specifically avoided it to prevent too much blocking of implied p-name
#
Calli
These are relevant test cases for whether the implied URL property would get a value from a different element or not:
#
Calli
TC1: <div class="h-x"><a href="alink"></a><area href="arealink" /></div> TC2: <div class="h-x"><a class="u-x" href="alink"></a><area href="arealink" /></div>
#
tantek
real world <area> handling in general is something we need to / are re-assessing
#
Calli
ok
#
tantek
there are use-cases for capturing the shapes of areas
#
tantek
e.g. photo notes
#
tantek
but we haven't quite figured out a good way of doing so
#
tantek
we tried one way, but found it didn't work in the micropub serialization scenario
#
tantek
so it is again an unsolved problem
#
tantek
Calli - if you have real world examples of <area> usage with microformats, or even real world examples of <area> usage that could *benefit* with addition of microformats, please provide the URLs as those would provide excellent emprical data to shape proposed solutions
#
Calli
I don't have any uses for area for my interests, I picked a test with A and AREA because of how the spec deals with only-of-type - needed an element that could be used forimplied URL prop that wasn't of the same type as A
#
kevinmarks
the implied name there is from hfeed not h-feed
#
kevinmarks
so that is a 1->2 mapping bug
gordonb and Calli joined the channel
#
Calli
oh right - yeah parser bug then
#
Calli
so i guess I don't know if I'll encounter h-feeds without names in the wild, but if I do...
#
kylewm
edited /microformats2-parsing-issues (+790) "/* implied properties when an explicit class is provided */ suggestion: split these out per property. I suspect it only affects u-url."
(view diff)
#
kevinmarks
I wonder if implied name should include children or not
#
kevinmarks
tricky as with h-feed you probably don't want that, but with h-review of an h-card, you may
#
kylewm
edited /microformats2-parsing-issues (-5) "/* implied properties when an explicit class is provided */ simplify suggested modification to implied url prasing"
(view diff)
#
tantek
oh hey welcome back Loqi wiki diffs!
#
Loqi
grins profusely
#
tantek
kevinmarks - yes, due to nested h-card or other structures, you do want that
gordonb and Calli joined the channel
#
Calli
kylewm makes good points about name and photo in his edit (namely that suggested changes won't really affect name because of the fallback to textContent and changes are possibly counterproductive for photo where it feels like presence of an image even if it's marked with another property is still strong enough evidence that the image is appropriate for the implied photo property)
#
tantek
yes that does seem sensible
#
tantek
I wonder if it is too much "smarts"
#
tantek
and will make it feel unpredictable
#
Calli
which leaves just implied URL
#
tantek
that's my only concern
#
tantek
but it is only a soft concern, I can be swayed
#
Calli
which then makes me think that for simplicity sticking with existing spec is good enough
#
Calli
Is it better to provide an incorrect URL or no URL at all? How would an author detect either case? Do any of the parsers focus on authoring scenarios? Do any of them highlight implied properties or missing properties?
#
Calli
So having considered it, I think I'm in the camp of leave the spec as it is
#
tantek
bad data is worse than no data
#
tantek
so incorrect URL is worse than no URL
#
tantek
in general
#
tantek
ok interesting
#
tantek
(re: camp of leaving spec as is)
#
tantek
that's very useful to know - please feel encouraged to note that in the issue!
#
Calli
I agree - but then my solution would be not to imply URL (or name or photo)
#
Calli
Seems like you're concerned with the authoring side of things and that's why you've gone for implied properties right? But how would an author be made aware that either they are missing a value for URL where it might be expected or they are getting a value that they might not expect?
#
tantek
all valid questions
#
tantek
that is one of the reasons that I proposed the relatively simple rule of - if the author has specified a u-* property on an element, then no other u-* properties are implied.
#
tantek
same with if the author has specified a p-* property, then no p-* properties are implied
#
tantek
on that element
#
tantek
that is a pretty simple rule to remember
#
tantek
by specifying properties on an element, you as an author signal that you want that element to mean/do precisely *this* (whatever those properties are), and *nothing* else.
#
Calli
yeah - that is a good rule.
#
tantek
the cross-interaction of "if you specify a p-* property then no u-* props are implied" and "if you specify a u-* property then no p-* properties are implied" works via a different simplification
#
kevinmarks
that is a good rule
#
tantek
that is
#
kevinmarks
the nesting issue is trickier
#
tantek
by specifying *any* property on an element, you as an author signal that you want that element to mean/do precisely *this* (whatever those properties are), and *nothing* else.
#
tantek
hence why the two proposals are listed there
#
tantek
in the issue
#
tantek
I wanted feedback / input on both
#
tantek
both are relatively simple rules, but different enough that to "new eyes" one may appear better than the other
#
tantek
thus really interested in seeing recorded feedback on both
#
tantek
thank you Calli kevinmarks !
#
Calli
On a related note: When considering parsing spec changes, do you (the community) have a large body of MF2 data that you don't want to change the meaning of? Or are changes still acceptable? If you're trying to avoid radical reinterpretations of existing data, do you have a crawler or some other means of getting hold of it?
#
tantek
we haven't gotten quite to that point yet - though there is a growing amount of indieweb data based on mf2
Calli joined the channel
#
Calli
More random thoughts about implied properties: Why doesn't implied name generate text content like an explicit p- property (which does converts imgs to their alt/src attributes insertion)? Is that difference intentional?
#
@fjdekermadec
Nobody cares, but I finally implemented #microformats on the sites of all our clients. 2010 @Google called and they want their index back!
(twitter.com/_/status/671525766564347904)
KartikPrabhu and tantek joined the channel
#
kevinmarks
the large body is probably in brid.gy/webmention.*
andicascadesf, tantek, KartikPrabhu1, nitot, tantek_, adactio, KartikPrabhu, Erkan_Yilmaz, eschnou, TallTed, KevinMarks_, MeanderingCode, KevinMarks, Left_Turn, kevinmarks_, gRegorLove, Calli, KevinMarks__ and kevinmarks joined the channel