hatom-issues

From Microformats Wiki
Revision as of 17:55, 3 January 2006 by FuzzyBSc (talk | contribs) (Add Published vs Updated commentary)
Jump to navigation Jump to search

Discussion Participants

Editors

Authors

Purpose

Questions or comments about hAtom go here. Please add your name to the Contributors section above.

Goals for hAtom

  1. to provide a blog-post microformat, based on how people actually produce weblogs
  2. based on (1), use Atom as it provides the most suitable data model for doing so
  3. based on (2), to make the format useful anywhere Atom might be used in context to create a syndication feed
  4. provide a baseline envelope format for similar {title|link|content|summary} web pages

Anti-Goals for hAtom

  1. Not to tell people how to write blogs or what there blog should look like; hAtom marked up blogs should look and behave identically to what they before hAtom was applied.

Questions and Comments

Limitations

There seems to be nothing in the hAtom specification to supply metadata for the blog (title, description, url, feedurl). There is nothing defined for the encapsulation of comments, comment counts, or links to comment sections. The microformat would be much more useful with these capabilities added. -- singpolyma 03:35, 3 Jan 2006 (PST)

Nomenclature

Note: microformats naming-principles include a precise means for coming up with names which should work in the 90% case.

One point to raise for hAtom in particular - we don't simply omit spaces from multiword property names, we use hyphens. E.g. "given-name" and many others in hCard.

Why atomentry?

class="atomentry" (or rather, "atom-entry")
Why not simply "entry"? The parallel to Atom is clear, but in the

context of a Web page, why add the reference? In case maybe you want to try for something approaching a string that won't get confused, my feeling is: forget it. Stick to the local semantics and let the doc-level (or HTML5 div level?) profile attribute disambiguate. Or to put it another way, it's premature to see a need at that point. -- DannyAyers

  • I (David Janes) chose the "atom" prefix:
    • to disambiguate; it is just too likely that "entry" or "feed" would appear on a random webpage in some other context. My preference would be to have a declarative statement in the XHTML header which would render this argument moot, but at this point the community seems cool on the concept.
    • to follow the naming pattern seen in the other "major" microformats (hCard, hCalendar, etc.)
    • because Entrys will not be required to be in Feeds (these rules and the reasons where this can happen will be forthcoming), I choose to disambiguate both
I don't like the analogy; I think this is more useful than just Atom, so it should be made generic. Dr. Ernie 16:59, 25 Oct 2005 (PDT)
DannyAyers My point exactly, but it wouldn't be the end of the world if the prefix was there - not reallly more than aesthetics...

STATUS - RESOLVED. We're going with "entry".

Tantek: This is actually difficult to consider outside the following issue. In particular, if "entry" is to serve as a potential root class name (similar to "vevent", which may be a root of an hCalendar event, or may be present in the context of a "vcalendar"), then we should strongly consider "uniquifying" it per our root-class-name practices. Possibilities to consider:
  • atom-entry
  • hentry
  • vjournal (from RFC 2445 and thus borrowed in effect from hCalendar)


Why atomfeed?

class="atomfeed" (or rather, "atom-entry")
As above on the atomprefix. But what does 'feed' mean in the context of a HTML page? Doesn't the <head> element cover the corresponding semantics?

-- DannyAyers

  • It is possible, somewhat common, and documented, that multiple feeds can appear on a single page, so it's insufficient to depend on the header, even though this may be the default case. You'll note that I've left out documenting a lot of concepts relating to feeds at a conceptual level, except for noting they exist because I think this is a bit of a swamp that's going to need more thinking
    • I'm going to more explicitly recognize that the XHTML document may act as an implicit feed in many cases
  • A Feed is a group of related Entries; what defines the relationship is entirely up to the author of the blog, except to note that if they were to place them together in the same Atom syndication feed, you'd do the same in the XHTML
This makes sense to me, the way vcalendar is optional since vevent is usually sufficient. Dr. Ernie 16:59, 25 Oct 2005 (PDT)
Ernie is precisely correct. The vevent/vcalendar :: entry/feed analogy is precisely correct. - Tantek
The multi-feed point makes sense, but if this data appears on a regular HTML page the question remains, does "feed" make sense? (Maybe just naming aesthetics again) -- DannyAyers
I'm thinking about it more -- I think so, just to split the content of the webpage up (as opposed to blogrolls, headers, footers, etc.) -- David Janes
Agreed with David. Not only does it make sense, it is a bad idea to consider renaming something like that for "aesthetics". - Tantek

STATUS - PARTIALLY RESOLVED. We're going with "feed" IF and when the Feed element is used. When and where Feed is used at all is still under discussion in the mailing list as of 2005-12-26.

Tantek: Per the root-class-name naming practices, we should seriously consider a more "unique" name, e.g. some possibilities:
  • atom-feed
  • hfeed

Why rel="link" ?

I know this maps through to the atom name, but rel="bookmark" is the established standard for permalinks, and is included in the w3c list of rel's, so there is an Occam's Razor case for using this. KevinMarks

  • I'd like input from everyone in this -- I'm torn really. Once I knock this thing into more of a complete state, I'll throw this out onto the mailing list for discussion -- DavidJanes
  • Also, "link" is horribly generic and is in fact modified through the "rel" attribute in Atom -- DavidJanes
  • Agreed with what Kevin wrote. Also, rel="link" doesn't actually make sense when you do the analysis as described in the rel-faq. The destination of the link is not really a "link" itself with respect to the current document/file. - Tantek
  • OK, I'm happy with this -- David Janes

STATUS - RESOLVED. We are using rel="bookmark".

  • Tantek: agreed.

title already defined by hCard

The title class is defined by hCard to mean "job title". Possible alternatives include (Please add to list):

  • 'summary', as used by hReview, hCalendar, VJOURNAL
    • Though I agree with the reuse, in this context, it may be confusing for those reading/familiar-with the Atom specification. We may want to avoid the use of 'summary' entirely within hAtom. - Tantek
  • 'Subject', as used by SMTP email
  • 'heading'
  • 'entry-title'

summary already defined and used by vCalendar/iCalendar/hCalendar/hReview

The summary class is defined by vCalendar, iCalendar, hCalendar, and also hReview, to mean "summary or title". Possible alternatives include (add to list):

  • description, as used by VJOURNAL. It may be possible to interpret description as text longer than summary which is about the entry content. The hierarchy of detail would be summary (atom:title) -> description (atom:summary) -> content (atom:content)
    • description is used ambiguously by RSS to mean 'content' or 'summary', and by hReview and hCalendar to mean 'content'. Doing this would recreate that ambiguity needlessly, when Atom distinguishes it clearly. Kevin Marks 15:51, 31 Dec 2005 (PST)
    • Kevin's right, and not only that, "description" does NOT mean summary in VJOURNAL. "description" means "full description" in vCalendar, iCalendar, hCalendar, and also hReview. See below for where to use "description". We must NOT use "description" to mean summary. - Tantek
  • Depending on your interpretation, atom:summary could still be summary. Here is another option: subject (atom:title) -> summary (atom:summary) -> content (atom:content).
  • atom:summary is described as abstract in

Tantek: We may want to avoid the use of 'summary' entirely within hAtom. Here are some alternatives:

  • excerpt - in practice this is what most people provide instead of full text.
    • +1 Tantek
  • partial-description
  • abstract - Rarely do bloggers actually provide abstracts different from the full text of their posts - this is so NOT the 80% (and thus should be rejected for microformat consideration), but see blog-post-examples for some examples.
  • content-summary

Why content?

The concept behind atom:content has precedent in earlier microformats derived from the iCalendar standard as "description".

  • Tantek: 'content' is a terrible name to use for this sort of thing. The term is far too ambiguous and overloaded, and nearly anything can be considered to be "content".
  • Using 'description' is well established by vCalendar, iCalendar (RFC 2445), hCalendar, xFolk, hReview etc. - Tantek

Proposals/alternatives:

  • description - re-use it as it is already used by vCalendar, iCalendar (RFC 2445), hCalendar, xFolk, hReview etc.
    • +1 Tantek

Date and time names alternatives

  • atom:updated
    • VJOURNAL LAST-MODIFIED
    • dtstamp
    • dtupdated
  • atom:published
    • VJOURNAL CREATED
    • dtpublished

Relationship to hReview definitions needs clarification

hAtom will define terminology for the general act of publication that overlaps with hReview's terminology for the specific act of publishing a review of something. The following terms could be pushed back into hReview:

  • atom:published -> hReview dtreviewed
  • atom:author -> hReview reviewer

"Pushed back" is the wrong direction here.

The right direction is "re-use" by new proposals/drafts. If you see anything in hReview that appears to overlap this new specification, the first thing to do is to see if you can reuse those terms from hReview in this new specification, not vice versa.

In addition, "published" does not mean the same as "dtreviewed" (you might write a restaurant review just after you eat there, but not actually "publish" it until later). "reviewer" is also a more precise semantic than "author", thus the two should not be collapsed.

- Tantek

hCards

Should hCards be required for the <address> of the Entry Poster? MAY, MUST, SHOULD? Your thoughts please -- DavidJanes

  • Robert Bachmann: “MUST” or at least “SHOULD” because atom:author is specified as “The "atom:author" element is a Person construct that indicates the author of the entry or feed.” and <address>’s semantics are too loose to describe an Atom person construct but using <addr class="vcard"> we would have pretty good 1:1 mappings:
    • atom:name ↔ hCard’s FN
    • atom:email ↔ hCard’s EMAIL
    • atom:uri ↔ hCard’s URI

STATUS - RESOLVED. "MAY" is the answer.

Tantek: I think this should be MUST. Atom should have referenced vCard for these semantics and made the mistake of making up their own terms. Let's undo that mistake with hAtom. Also, hReview 0.3 is going to make hCard a MUST for the "reviewer" property, based on experience and feedback. Thus we may want to just follow suit with hAtom as well.

Comparisons

This seems precisely analogous to S5:

  • atomentry <-> slide
  • content <-> slidecontent
  • summary <-> handout

I'm all for NOT boiling the ocean, but these really seem like the same cup of tea.

-- Ernie Prabhakar

  • See the #Purpose section above. Basically that drove the design decision for the naming David Janes

STATUS - RESOLVED. We're sticking with atom terminology (entry, content, summary).

See above as David says. The atom terminology is both problematic, and doesn't make sufficient reuse of existing microformat terminology. As far as the analogy to S5, yes, there is an analogy, but that doesn't make them the same. The semantics that are represented are different enough to let these evolve independently and see if content authors want them to converge or not. Note that you can overlay hAtom and S5 in the same markup. Anyone that is serious about converging these should *try* using both at the same time in a *real* slide presentation example and report back their experience. - Tantek

Repeated Elements

We allow certain elements to be repeated, such as Entry Permalink, Entry Published and Entry Title, even though there can be at most one real value. We provide "disambiguation" rules for sorting out which is the real value. See here, here, here and here.

Your thoughts please... -- David Janes

STATUS - RESOLVED. The spec has explicit rules for disambiguating all these items if they appear multiple times.

Opaqueness

If you have concerns about opaqueness, that is, stopping interpretation below certain hAtom elements, raise them here.

Opaqueness of other microformat elements

How would we handle a case where someone wanted to provide a vcard under the class=entry element for an individual who was neither author or contributor? Consider the hypothetical case where someone wanted to list their "muse" alongside article author and contributors. If this vcard included a title it might be included accidentally as an <atom:title>.

To summarise, Is it possible that other microformats found under the class=entry or class=feed elements need to be considered opaque?

-- BenjaminCarlyle

  • The issue of "muse" and such is somewhat out of scope. However, I grasp your larger point -- what if we wanted to extend or compositie hAtom in the future. Given the 80-20 rule right now, my feeling is to set aside the problem and if it arises, define a class="opaque" element. -- David Janes

See the mfo-examples document, and add further thoughts on this matter there. -- Tantek

Opaqueness of summary and content

What one publisher considers the entry content may differ from another publisher's point of view. Is the content simply a div that does not contain any author/updated/published metadata etc, or could some of that metadata be relevant to the content as well as the entry? Consider updated. last-modified-brainstorming introduces an idea of using <ins> and <del> elements to indicate update time. Updates are also often included in entry content with further information. This suggests to me that the line of opaqueness is blurry.

Perhaps content and summary should not be opaque, and instead rely on the mfo proposal to avoid parsing into microformats below the content level. This approach would allow a single div to contain both "entry" and "content" classes should all metadata be considered content by the author, or would permit any other subset of the metadata to be considered content without repeating one's self.

Consider also the "read more"-style blog. The following nesting of div elements is illegal under current opacity rules:

<div class="content"><div class="summary">......

A further example is provided by _fil_ on #microformats, who uses the rel-tag microformat within his atom:content to be handled as tags in his feed reader.

Identification

The current spec under Schema:Nomenclature:Entry includes the text: "if practical, also define id="unique-identifier" to the Entry" What should be done with this id by parsers? How does this interact (if at all) with the interpretation of a rel=bookmark within the entry?

Also, how should a feed <id> element be filled out from a hAtom source document? Is a rel=bookmark at the feed level required?

The id elements in atom are supposed to survive all future movements of the blog t new hosting arragements and the like. Are current feed URLs or even rel=bookmarks solid enough?

STATUS - OPEN.

HTML Title

Atom permits title to be either plain text or html. hAtom2Atom.xsl currently uses a plain text translation, and some feed readers seem not to handle html titles well (liferea does not normalize-whitespace, for example). Should a hAtom title element become a plain text or a html atom title? If so, should a subset of html be passed through rather than all html (including id, etc)?

rel-tag

Should hAtom use rel-tag for atom category elements?

IMHO yes. -- Tantek

A version of this is currently implemented in hAtom2Atom.xsl, but the interpretation of rel-tag is not straightforward.

rel-tag uses the last path segment of a URI as its tag, for example <a href="http://apple.com/ipod" rel="tag">iPod</a>. Human-friendly content is permitted within the anchor. Atom defines three attributes on a category element. "term" is the category in use. "scheme" is a namespace for this category. "label" is a human-friendly text-only version of the category.

  • This looks like a clear mapping to me - term is last path segment; scheme is the tagspace and label is the text within the anchor? The problem is if the scheme + tag is not a true URL but a URI. So for your example, term is 'ipod, scheme is 'http://apple.com/' and label is iPod. Kevin Marks 15:03, 31 Dec 2005 (PST)

hAtom2Atom.xsl does not currently supply a scheme. Label is taken from the content of the anchor tag, and no special handling for content such as the title attribute of an img element is performed. Term is the portion of the href after the last slash character.

rel-tag permits url encoding for IRIs, as well as conversion of spaces to plus (+) characters. It is unclear whether the conversion of rel-tag data to atom:category/@term should attempt to reverse any such encoding. The handling of plus characters may be especially difficult to reverse (are the plus characters, or spaces?).

  • They are spaces. If you want plus characters use %2B Perhaps I should add this to rel-tag. Kevin Marks 15:03, 31 Dec 2005 (PST)

Excess disambiguation rules?

Disambiguation rules apply to feed and entry title, and hAtom2Atom.xsl implements these. Rules also apply to permalink, published, and updated. These are currently not implemented. If they appear multiple times in the source document they are repeated multiple times.

It is clear that the data relating to these fields may be repeated within a hAtom entry, however the class notation may not. Only one element need be marked with rel="bookmark". Only one need be marked published, and one updated. Should the disambiguation rules be removed and only one element be allowed for each value, or is there value to the publisher in marking different elements with the hAtom class names?

Dependancies

mfo

Does this specification depend on acceptance of a hAtom-compatible mfo? See mfo-examples.

last-modified

Does this specification depend on acceptance of a hAtom-compatible last-modified?

Is atom:content necessary?

Atom's structure is built up around separating content and other metadata. atom:updated, atom:author, and the like are separate from atom:content any may contain repeated data. Microformats are built around bringing the content and the metadata back together. Is there are genuine use case for identifying only part of the atom entry as content? Presumably the whole html entry is fit for human consumption, or it wouldn't be part of a microformatted web page. Could that whole html snippet be used as the content?

Published as default value for atom:updated

It seems to be common practice to include an "updated" section within the main blog content to track updates to an atom:entry as they occur. It is less common to include a value for atom:published within atom:content. atom:published is usually provided by a machine, but atom:updated is often provided by a human.

I suggest that if a value of published exists but no value for updated exists that the required updated field be filled out from the optional published field. I think this would make changing the required value of updated easier for publishers. Also, several updates may occur to a single entry. I suggest that a disambiguation rule be applied such that the the latest timestamp of any updated field be used if several exist. The overal parser semantics would therefore be:

  1. If multiple updated fields exist, choose the most recent one.
  2. If only one updated field exists, choose that value.
  3. If no updated field exists but a published field exists, use the published value for atom:updated.

See Also