Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On Feb 7, 2007, at 10:30 PM, Derrick Lyndon Pallas wrote: Yes, there are other ways to solve the problem; in fact, I do solve the problem in an unelegant way. My real issue now is (as laid out above) the resistance to real discussion of the problem. I think what you're seeing is that microformats value publishers above parsers. Anything that makes publishing even marginally more difficult to make parsing easier is a non-starter. As long as parsing is *possible* (and you seem to agree it is), that's good enough for microformats. The theory is if we can make it absolutely as easy as possible for publishers, they'll flood the web with semantic content, and parsers will work through the difficulty to get that content. And you'll find plenty of interest here in helping through the difficulty. But not much in shifting any (not even a bit) of the difficulty onto publishers. That's just fundamentally not how microformats are designed. If you want a general means of consuming microformat content, just consume RDF. GRDDL [1] converts specific microformats to RDF, so any RDF consumer is effectively a generic microformats consumer. When new microformats are developed, they'll end up in RDF as soon as someone works out the GRDDL, and you won't need to update your RDF consumer at all. [1] http://www.w3.org/2004/01/rdxh/spec Peace, Scott ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/7/07, Ryan King [EMAIL PROTECTED] wrote: On Jan 31, 2007, at 7:07 PM, Derrick Lyndon Pallas wrote: If I have a parser that only knows (and only cares about) the rel- tag format, it will be confused by people that use rel-tag for the category property in hCard. It seems unreasonable that every microformat should have to know about every other microformat, especially when they are nested. Actually I think it *is* quite reasonable to make parsers know about every microformat. Microformats are designed to be easy to publish, even when that means that they're hard to parse. Simple economics show that it's much more valuable to make publishing low-cost, because the increased in published data will allow you to amortize the cost of writing and maintaining parsers across more transactions. Also, microformats are not designed to be generic or open ended, but specific solutions to specific problems. Requiring authors to add markup in order to make rel-tag's scope explicit makes it hard to publish the data and doesn't solve any real problem. No one has said anything about any required mechanism, except for the unreasonableness of requiring a parser to know about every possible semantic available for a given encoding. The point is that the rel-tag spec lays out a contract: http://microformats.org/wiki/rel-tag#Scope Scope rel=tag is specifically designed for tagging content, typically web pages (or portions thereof, like blog posts). ...that seems to contradict itself... If you need to define tags as part of a more specialised format, rel=tag is the recommended way to do so, and xFolk, hReview, hCard and hCalendar all do this. The assumption here seems to be that when a microformat appears on a page, the subject matter of the microformat is highly correlated with what the page is about (whatever that means...). The justification given for this assumption is the historical record for real publishing, and by concrete example, the book publishing industry. If this assumption is true, then the subject of the rel-tag is largely a non-issue. It's hard to swallow this assumption, because it seems possible, and highly likely, that authors' intentions are different. As an author, it would surprise me to learn that the categories I specify to describe my friend would also be understood by a parser to describe the page itself. This creates a mismatch between the functional model and the mental model of the author, and possibly other human consumers as well. (Human consumers aren't going to consider the page a subject of contained rel-tags, either...) When I explicitly publish a piece of information, I expect it to be interpreted within that context. When other behaviour creeps in, the integrity of my authorship becomes brittle. I agree with Ryan that this is an authorship issue, and disagree that it's not an issue. Authors should know that any rel-tag useage could be applied to the page as a whole. This isn't made clear in the formats that reuse rel-tag. Currently, rel-tag is difficult to use as a publisher, because the subject the tag applies to has become modal. Either we need to remove the modal nature of the subject, make it explicit, or provide a mechanism to control it. -Ben ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On Jan 31, 2007, at 7:07 PM, Derrick Lyndon Pallas wrote: If I have a parser that only knows (and only cares about) the rel- tag format, it will be confused by people that use rel-tag for the category property in hCard. It seems unreasonable that every microformat should have to know about every other microformat, especially when they are nested. Actually I think it *is* quite reasonable to make parsers know about every microformat. Microformats are designed to be easy to publish, even when that means that they're hard to parse. Simple economics show that it's much more valuable to make publishing low-cost, because the increased in published data will allow you to amortize the cost of writing and maintaining parsers across more transactions. Also, microformats are not designed to be generic or open ended, but specific solutions to specific problems. Requiring authors to add markup in order to make rel-tag's scope explicit makes it hard to publish the data and doesn't solve any real problem. -ryan -- Ryan King [EMAIL PROTECTED] ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
Ryan King wrote: Actually I think it *is* quite reasonable to make parsers know about every microformat. This is not viable from a consumer perspective. New formats can immediately invalidate old parsers by changing the semantics the consumer expects without so much as an annotation in the definition of the affected format. (Incidentally, this is the same sort of problem Aspect Orientation has.) The decoupling of a format's semantics from the format definition has the additional effect that users may be struck by unexpected semantics. Formats like rel-license suffer from the same problem. Microformats are designed to be easy to publish, even when that means that they're hard to parse. Simple economics show that it's much more valuable to make publishing low-cost, because the increased in published data will allow you to amortize the cost of writing and maintaining parsers across more transactions. This is a straw-man. It doesn't make them harder to publish or add to cost by adding the meta uf (or scope or whatever) to @class for top level formats, especially to ask users (or generators) who are USING a format to mark where the format begins because (presumably) they understand that particular format. Also, microformats are not designed to be generic or open ended, but specific solutions to specific problems. Certain features (like rel-tag or rel-license or rel-*) ARE being reused, in practice; it is bad engineering to limit their usefulness. There is a problem; I know because I'm a consumer. When an issue comes up every three months and is brushed off as not an issue every time, that is dishonest. Potential consumers have found it to be a problem in practice; and yet, the current consumers think it is not a problem because they don't see any need. Limiting the usefulness of something prevents results in the wild, which stunts future progress. Requiring authors to add markup in order to make rel-tag's scope explicit makes it hard to publish the data and doesn't solve any real problem. Again, straw-man. Changing a string from vcard to vcard uf or xfolkentry to xfolkentry uf is NOT HARD for the author of a generator. On the other hand, it is much harder for a parser to magically know all the current (and future) microformats. It does solve real problems. Yes, there are other ways to solve the problem; in fact, I do solve the problem in an unelegant way. My real issue now is (as laid out above) the resistance to real discussion of the problem. ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/2/07, Derrick Lyndon Pallas [EMAIL PROTECTED] wrote: Except it does need it. Say you put your del.icio.us (or otherwise) feed on your page and want to include it and the associated tags as xFolk entries. How can a generic rel-tag parser know that the xFolk entires don't apply to the current page without knowing about xFolk. That's the scoping problem. The tag applying to the page just means that there's something on the page relevant to that tag. And there is - the del.icio.us feed! The problem is not that they may be applied to the page it's that they are applied to the page I meant 'may' as in 'yes, the parser can go ahead and apply them' - my ambiguity sorry. and there are reasons that is inappropriate, Can you expand on the reasons? Basically, if a page has a blog entry about Cats and an hCard in the category 'Dogs' on it, why can't that page validly be tagged with 'cats' and 'dogs'? My solution (to indicate scope with a generic rel-tag counterpart and then allow specific parsers to override the scoping rule if they understand the containing element) is both general and powerful. I haven't looked at the different scoping proposals and certainly I'm not saying yours is bad, I'm questioning the need to complicate what is after all an incredibly simple format. Take the example of a dead relative: there is no way to put a family tree with relatives you need to tag as deceased on your own page without a document level parser concluding that you are dead. That doesn't make any sense to me. All a rel-tag parser would take from it would be that the page had something on it about someone who's 'dead', surely. I don't know where it starts making inferences about me. -Ciaran ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/2/07, Derrick Lyndon Pallas [EMAIL PROTECTED] wrote: Take the example of a dead relative: there is no way to put a family tree with relatives you need to tag as deceased on your own page without a document level parser concluding that you are dead. --- that is not true, you are not confusing that YOU ARE DEAD, you are only saying that the page has information about 'dead'. Rel-tag when applied to the whole page simply says that there is some information on this page related to 'X'. Nothing more... when you scope it to a specific microformat it gains further meaning about ONLY that object. In your family tree example, a rel-tag crawler would find a rel-tag of dead on the page and index it under the tagspace of 'dead'. This is expected, perfectly valid, and correct behavior. Now, when you look to a specific hCard searcher/spider, it will NOT apply rel-tag of 'dead' to all the hCards on the page, ONLY to the ones where it has been scoped (this is done my adding the rel-tag inside the class=vcard) it will not mix-up things and assume you are dead. Does that make sense? -brian -- brian suda http://suda.co.uk ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
Brian Suda wrote: Does that make sense? That's what I get for using someone else's example. Still, no one has responded to the more fundamental concern that rel-tag is not reusable for things like lists of bookmarks. (Or does someone really find it helpful that a page has content about X can just mean that a page has a bookmark to X; or at least it did when I indexed the page, I guess its not there anymore because it rolled off.) ~D ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
Ciaran McNulty wrote: The tag applying to the page just means that there's something on the page relevant to that tag. And there is - the del.icio.us feed! The tag applies to the link; not the content, and certainly not the whole contents of the page. If I search for pages with tag foo and a page with a link to something tagged foo comes up, that's not what I wanted. Can you expand on the reasons? Basically, if a page has a blog entry about Cats and an hCard in the category 'Dogs' on it, why can't that page validly be tagged with 'cats' and 'dogs'? It can be. But xFolk and hReview, etc., specifically change the semantics of rel-tag in their definitions. The problem is that there is no way to tell where that semantic shift ends (i.e. what scope it has) without understanding xFolk and hReview. Are we going to require all old microformat parsers to understand all new microformats? I haven't looked at the different scoping proposals and certainly I'm not saying yours is bad, I'm questioning the need to complicate what is after all an incredibly simple format. Consistency - the design must not be inconsistent. A design is allowed to be slightly less simple and less complete to avoid inconsistency. Consistency is as important as correctness. http://www.jwz.org/doc/worse-is-better.html ~D ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/2/07, Derrick Lyndon Pallas [EMAIL PROTECTED] wrote: Still, no one has responded to the more fundamental concern that rel-tag is not reusable for things like lists of bookmarks. --- i think the issue might be with what you WANT rel-tag to be and what it is? at the moment rel-tag would say this page as content about X. That is ALL rel-tag knows about and that is all a rel-tag spider is concerned with. (Or does someone really find it helpful that a page has content about X can just mean that a page has a bookmark to X; or at least it did when I indexed the page, I guess its not there anymore because it rolled off.) ~D It doesn't matter if it is a bookmark, blog post, hAtom entry, hCalendar or hResume, etc. A rel-tag applies to the page. Organizations like Icerocket, Technorati and others index TAGS all they are conserned with is getting you to the data. (freshness of that data is another problem, sure things might roll off, but things might go 404 too! that's life) I'm not sure what you mean when you say is not reusable for things like lists of bookmarks. why not? when a rel-tag spider comes around and finds the tags it gets indexed in one way. When a bookmark/xFolk spider comes along it indexes things differently. I get the feeling you want both the rel-tag and bookmark spiders to index it in the exact same way? At the moment this is NOT how rel-tag works. Do you have a specific use-case or URL you want us to look at? otherwise we should stay away from hypothetical what ifs. -brian -- brian suda http://suda.co.uk ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
Brian Suda wrote: I get the feeling you want both the rel-tag and bookmark spiders to index it in the exact same way? At the moment this is NOT how rel-tag works. Do you have a specific use-case or URL you want us to look at? otherwise we should stay away from hypothetical what ifs. No, I don't want those spiders to behave the same way and I don't think I've been unclear on my reasoning behind this or my use-case, which has been presented several times. I'd rather not have wild guesses; therefore: I, as a consumer of rel-tags, would like a way to know when the specificity of a rel-tag changes without having to write down and check a list of formats that change it. I'm not suggesting that we change the semantics of rel-tag, I'm suggesting that we add a marker that indicates where other microformats begin so we can tell when the semantics of features change. At any rate, Tantek said today that it is a non-issue that comes up quite often, so at his request, I put a section on http://www.microformats.org/wiki/rel-tag-faq describing his explanation applied to my use-case, that of blogrolls. And for the record: yes, I do understand the difference between normative and positive specification. ~D ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/1/07, Derrick Lyndon Pallas [EMAIL PROTECTED] wrote: If I have a parser that only knows (and only cares about) the rel-tag format, it will be confused by people that use rel-tag for the category property in hCard. It seems unreasonable that every microformat should have to know about every other microformat, especially when they are nested. Can I ask what the confusion is? If I have a hcard with a rel-tag indicating 'football' in that hCard, then the naive interpretation that 'this page has something on it to do with football' that your parser will take from it is probably correct. -Ciaran McNulty ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
Ciaran McNulty wrote: Can I ask what the confusion is? If I have a hcard with a rel-tag indicating 'football' in that hCard, then the naive interpretation that 'this page has something on it to do with football' that your parser will take from it is probably correct. What about an xFolk link with a tag of http://wikipedia.org/wiki/NSFW? Should that imply that the containing page is not safe for work? Do the xFolk entries on unalog imply that unalog is about any of those tags? Here's my problem: rel-tag is reusable. It applies to whatever contains it. Well, except under specific circumstances which are documented in the other formats in which it has been reused, then it only applies to a sub-container, which we didn't mark in a generic way. I'm just looking for a generic scoping mark. ~D ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
Re: [uf-discuss] Should microformat features (like rel-tag) have explicit scope?
On 2/1/07, Derrick Lyndon Pallas [EMAIL PROTECTED] wrote: What about an xFolk link with a tag of http://wikipedia.org/wiki/NSFW? Should that imply that the containing page is not safe for work? Well if an item on a page is tagged NSFW doesn't that mean the page is NSFW? I must confess I'm not 100% familiar with xFolk. rel-tag is reusable. It applies to whatever contains it. Well, except under specific circumstances which are documented in the other formats in which it has been reused, then it only applies to a sub-container, which we didn't mark in a generic way. I'm just looking for a generic scoping mark. ~D My point is that rel-tag doesn't have any scope, and I'm sort-of arguing it doesn't need it. Take the example of a page that contains: * An hAtom entry tagged with 'FOO' * An hCard with the category 'BAR' An hAtom parser will correctly note that the only rel-tag in the hAtom entry is 'FOO' and so that's the category for the entry. An hCard parser will note that the only rel-tag inside the hCard is 'BAR, and so that category applies to the card. However, a generic rel-tag parser doesn't need to know don't look inside hAtom and hCard, as you seem to be suggesting. Any rel-tags it finds may be applied to the page itself quite fairly, and so a rel-tag parser would say 'this page contains something relevant to FOO and something relevant to BAR. Does that make sense? -Ciaran McNulty ___ microformats-discuss mailing list microformats-discuss@microformats.org http://microformats.org/mailman/listinfo/microformats-discuss
