Discussion:
[whatwg] rel=bookmark
Ed Summers
2017-08-06 01:06:57 UTC
Permalink
Hi all,

I was wondering if anyone can provide any information, or a pointer to previous discussion, about why the bookmark link relation can't be used with the <link> element [1].

The topic has come up recently on the IETF link-relations discussion list [2] where a new link relation has been proposed to encourage persistent linking [3]. The proposed 'identifier' relation seems to closely resemble the idea of a permalink (a persistent link) that can be found in the definition of bookmark. If bookmark allowed use with the <link> element then I think there would be less of a demonstrated need for the new 'identifier' link relation.

Thanks for any information you can provide. I apologize if I'm restarting a conversation that has already happened.

//Ed

[1] https://www.w3.org/TR/html5/links.html#link-type-bookmark
[2] https://www.ietf.org/mail-archive/web/link-relations/current/msg00670.html
[3] https://datatracker.ietf.org/doc/draft-vandesompel-identifier/
Domenic Denicola
2017-08-06 01:19:28 UTC
Permalink
Hi Ed,

(Remember to use the HTML Standard, located at https://html.spec.whatwg.org/multipage/links.html#link-type-bookmark, not any forks of it.)
The bookmark keyword gives a permalink for the nearest ancestor article element of the linking element in question, or of the section the linking element is most closely associated with, if there are no ancestor article elements.
Your proposal is essentially to give it an entirely separate meaning when used in the context of the <link> element, but that's not usually how we share link relations between the different elements: cf. alternate, author, help, license, next, etc.

At least, that is how I understand; I'm having a hard time distinguishing what "identifier" is for in practice, and in particular why it is different than "canonical".

-----Original Message-----
From: whatwg [mailto:whatwg-***@lists.whatwg.org] On Behalf Of Ed Summers
Sent: Saturday, August 5, 2017 21:07
To: ***@whatwg.org
Subject: [whatwg] rel=bookmark

Hi all,

I was wondering if anyone can provide any information, or a pointer to previous discussion, about why the bookmark link relation can't be used with the <link> element [1].

The topic has come up recently on the IETF link-relations discussion list [2] where a new link relation has been proposed to encourage persistent linking [3]. The proposed 'identifier' relation seems to closely resemble the idea of a permalink (a persistent link) that can be found in the definition of bookmark. If bookmark allowed use with the <link> element then I think there would be less of a demonstrated need for the new 'identifier' link relation.

Thanks for any information you can provide. I apologize if I'm restarting a conversation that has already happened.

//Ed

[1] https://www.w3.org/TR/html5/links.html#link-type-bookmark
[2] https://www.ietf.org/mail-archive/web/link-relations/current/msg00670.html
[3] https://datatracker.ietf.or
Ed Summers
2017-08-07 14:27:07 UTC
Permalink
Hi Domenic,
Post by Domenic Denicola
(Remember to use the HTML Standard, located at https://html.spec.whatwg.org/multipage/links.html#link-type-bookmark, not any forks of it.)
Oops, my bad! Luckily the definition looks the same so I think my question is still relevant?
Post by Domenic Denicola
The bookmark keyword gives a permalink for the nearest ancestor article element of the linking element in question, or of the section the linking element is most closely associated with, if there are no ancestor article elements.
Your proposal is essentially to give it an entirely separate meaning when used in the context of the <link> element, but that's not usually how we share link relations between the different elements: cf. alternate, author, help, license, next, etc.
I don't think allowing rel=bookmark to be used with <link> would change the meaning because of the clause "... or of the section of the linking element is most closely associated with". If the <link> were used in the <head> of an HTML document then the bookmark would be associated with the HTML document itself, not some section within it.
Post by Domenic Denicola
At least, that is how I understand; I'm having a hard time distinguishing what "identifier" is for in practice, and in particular why it is different than "canonical".
Yes, I initially thought canonical was the logical choice too. But as the draft authors point out, canonical [1] says nothing about persistence and is used instead to indicate a preferred URL for duplicative content.

//Ed

[1] https://tools.ietf.org/html/rfc6596
Kevin Marks
2017-08-06 01:19:54 UTC
Permalink
That use case sounds more like rel="canonical"
Post by Ed Summers
Hi all,
I was wondering if anyone can provide any information, or a pointer to
previous discussion, about why the bookmark link relation can't be used
with the <link> element [1].
The topic has come up recently on the IETF link-relations discussion list
[2] where a new link relation has been proposed to encourage persistent
linking [3]. The proposed 'identifier' relation seems to closely resemble
the idea of a permalink (a persistent link) that can be found in the
definition of bookmark. If bookmark allowed use with the <link> element
then I think there would be less of a demonstrated need for the new
'identifier' link relation.
Thanks for any information you can provide. I apologize if I'm restarting
a conversation that has already happened.
//Ed
[1] https://www.w3.org/TR/html5/links.html#link-type-bookmark
[2] https://www.ietf.org/mail-archive/web/link-relations/
current/msg00670.html
[3] https://datatracker.ietf.org/doc/draft-vandesompel-identifier/
Philipp Serafin
2017-08-06 10:13:02 UTC
Permalink
As the IETF usecase seems to be about permalinks, is there any requirement
for rel=canonical regarding validity in the future?
Post by Kevin Marks
That use case sounds more like rel="canonical"
Post by Ed Summers
Hi all,
I was wondering if anyone can provide any information, or a pointer to
previous discussion, about why the bookmark link relation can't be used
with the <link> element [1].
The topic has come up recently on the IETF link-relations discussion list
[2] where a new link relation has been proposed to encourage persistent
linking [3]. The proposed 'identifier' relation seems to closely resemble
the idea of a permalink (a persistent link) that can be found in the
definition of bookmark. If bookmark allowed use with the <link> element
then I think there would be less of a demonstrated need for the new
'identifier' link relation.
Thanks for any information you can provide. I apologize if I'm restarting
a conversation that has already happened.
//Ed
[1] https://www.w3.org/TR/html5/links.html#link-type-bookmark
[2] https://www.ietf.org/mail-archive/web/link-relations/
current/msg00670.html
[3] https://datatracker.ietf.org/doc/draft-vandesompel-identifier/
Ed Summers
2017-08-07 14:06:35 UTC
Permalink
Hi Phil,
Post by Philipp Serafin
As the IETF usecase seems to be about permalinks, is there any requirement
for rel=canonical regarding validity in the future?
Yes, the quality of persistence is why I thought rel=bookmark worked best, although canonical was the relation I first thought of too.

As the IETF draft authors describe in a related blog post [1] canonical was dropped from consideration because it exists to "identify content that is either duplicative or a superset of the content at the context (referring) IRI" and does not speak to the durability of the link.

//Ed

[1] http://ws-dl.blogspot.com/2016/11/2016-11-07-linking-to-persistent.html
Ed Summers
2017-08-08 16:58:48 UTC
Permalink
Hi Kevin,
Post by Kevin Marks
That use case sounds more like rel="canonical"
You weren't the only one (myself included) who thought that. Michael Nelson, one of the authors if the identifier I-D, just wrote a blog post explaining why not canonical:

http://ws-dl.blogspot.com/2017/08/2017-08-07-relcanonical-does-not-mean.html

I think I'm convinced that canonical isn't the right fit for what they are talking about. But if rel=bookmark could be used in <link> elements I think it would work better than a slightly similar, oddly named, link relation, which IMHO is bound to cause confusion for web publishers.

//Ed
Kevin Marks
2017-08-08 18:01:48 UTC
Permalink
This sounds like what we use uid for in microformats - the url that you
want as the persistent identifier.

http://microformats.org/wiki/uid - it looks like you wrote this up a while
back, Ed.

See u-uid in h-entry http://microformats.org/wiki/h-entry
Post by Ed Summers
Hi Kevin,
Post by Kevin Marks
That use case sounds more like rel="canonical"
You weren't the only one (myself included) who thought that. Michael
Nelson, one of the authors if the identifier I-D, just wrote a blog post
http://ws-dl.blogspot.com/2017/08/2017-08-07-
relcanonical-does-not-mean.html
I think I'm convinced that canonical isn't the right fit for what they are
talking about. But if rel=bookmark could be used in <link> elements I think
it would work better than a slightly similar, oddly named, link relation,
which IMHO is bound to cause confusion for web publishers.
//Ed
Kevin Marks
2017-08-08 18:04:57 UTC
Permalink
See also http://microformats.org/wiki/sharelink-formats for a (recent)
related use case
Post by Kevin Marks
This sounds like what we use uid for in microformats - the url that you
want as the persistent identifier.
http://microformats.org/wiki/uid - it looks like you wrote this up a
while back, Ed.
See u-uid in h-entry http://microformats.org/wiki/h-entry
Post by Ed Summers
Hi Kevin,
Post by Kevin Marks
That use case sounds more like rel="canonical"
You weren't the only one (myself included) who thought that. Michael
Nelson, one of the authors if the identifier I-D, just wrote a blog post
http://ws-dl.blogspot.com/2017/08/2017-08-07-relcanonical-
does-not-mean.html
I think I'm convinced that canonical isn't the right fit for what they
are talking about. But if rel=bookmark could be used in <link> elements I
think it would work better than a slightly similar, oddly named, link
relation, which IMHO is bound to cause confusion for web publishers.
//Ed
Ed Summers
2017-08-08 19:43:40 UTC
Permalink
Post by Kevin Marks
See also http://microformats.org/wiki/sharelink-formats for a (recent)
related use case
Post by Kevin Marks
This sounds like what we use uid for in microformats - the url that you
want as the persistent identifier.
http://microformats.org/wiki/uid - it looks like you wrote this up a
while back, Ed.
Oh boy, that is going back a ways yes :) I see some of that documentation still refers to HTML 4!

I guess I'll put a contribution together that adjusts rel="bookmark" and see how it fares. Thanks for the feedback everyone.

//Ed
Ed Summers
2017-08-08 21:39:45 UTC
Permalink
Post by Ed Summers
I guess I'll put a contribution together that adjusts rel="bookmark" and see how it fares. Thanks for the feedback everyone.
I started with an issue ticket [1] that references this conversation in case anyone is interested in following along there.

//Ed

[1] https://github.com/whatwg/html/issues/2899

Continue reading on narkive:
Loading...