February 12, 2004
@ 11:47 PM
Comments [21]
It Begins

According to Jeremy Zawodney the “My Yahoo's RSS module also groks Atom. It was added last night. It took about a half hour.” Seeing that he said it took only 30 minutes to implement this and there are a couple of things about ATOM that require a little thinking about it even if all you are interested in is titles and dates as My Yahoo! is I decided to give it a try and subscribe to Mark Pilgrim's Atom feed and this is what I ended up being shown in My Yahoo!

dive into mark  Remove

The first minor issue is that the posts aren't sorted chronologically but that isn't particularly interesting. What is interesting is if you go to the article entitled The myth of RSS compatibility its publication date is said to be “Wednesday, February 4, 2004” which is about a week ago and if you go to the post entitled  Universal Feed Parser 3.0 beta its publication date is said to be Wednesday, February 1, 2004 which is almost 2 weeks ago not a day ago like Yahoo! claims.  

The simple answer to the confusion can be gleaned from Mark's ATOM feed, that particular entry has a <modified> date of 2004-02-11T16:17:08Z, an <issued> date of 2004-02-01T18:38:15-05:00 and a <created> date of 2004-02-01T23:38:15Z. My Yahoo! is choosing to key the freshness of article of its modified date even though when one gets to the actual content it seems much older.

It is quite interesting to see how just one concept [how old is this article?] can lead to some confusion between the end user of a news aggregator and the content publisher. I also suspect that My Yahoo! could be similarly confused by the various issues with escaping content in Atom when processing titles but since I don't have access to a web server I can't test some of my theories.

I tend to wonder whether the various content producers creating Atom feeds will ditch their feeds for Atom 0.4, Atom 0.5 up until it becomes a final IETF spec or whether they'll keep parallel versions of these feeds so Atom 0.3 continues to live in perpetuity.

It's amazing how geeks can turn the simplest things into such a mess. I'm definitely going to sit it out until the IETF Atom 1.0 syndication format spec before spending any time working on this for RSS Bandit.


 

Categories: Technology

« What is Metacrap? | Home | Sex, Lies and XML MIME Types »
Friday, 13 February 2004 02:09:46 (GMT Standard Time, UTC+00:00)
I'm not sure what your point is. Both the created and modified dates are correct on all my entries; I have been updating the Feed Parser Beta announcement with links to new versions as they come out, but other entries have not been modified. The data is correct and semantically unambiguous; what Yahoo chooses to do with it is entirely up to them.
Mark
Friday, 13 February 2004 02:11:31 (GMT Standard Time, UTC+00:00)
BTW, you can see the revision history of the Feed Parser entry here (this is linked from the bottom of the entry):

http://diveintomark.org/cgi-bin/history.cgi?id=3232
Mark
Friday, 13 February 2004 03:00:32 (GMT Standard Time, UTC+00:00)
Mark,
A lot of people think ATOM is RSS with a few tag names changed. These people will write sloppy applications that will confuse end users or simply won't work outside of trivial cases. Pointing to specs and claiming that you have moral superiority will not change this fact.

The fact is that most RSS consumers assume that there is one date attached to an item, Atom has three. Different apps will expose this to end users in different ways. Hilarity or confusion will ensue.

PS: Do you actually read my blog or do you just Feedster for your name?
Dare Obasanjo
Friday, 13 February 2004 03:19:03 (GMT Standard Time, UTC+00:00)
OK, ignoring for the moment that I didn't do *any* of the things you just mentioned, I'm really not sure what more you want out of an XML data format than tagged data that is correct and semantically unambiguous.

As to your point that no one should ever present more information than RSS does today because it would confuse people who have written their apps to work around the limitations RSS has today, well, what does the phrase "horseless carriage" mean to you?

I actually read your blog intermittently.
Mark
Friday, 13 February 2004 14:25:33 (GMT Standard Time, UTC+00:00)
Mark,
I had two points, the first is that the current spec is focused too much on content producers and not on content consumers which has lead to the introduction of concepts and features that don't jibe with the current aggregator landscape [content can be any MIME type, entries have 3 dates, content rel="fragment" vs. summary]. In the long term aggregators will probably adapt and realize in some cases, Atom isn't just RSS with some elements renamed but in the short term there will be some confusion.

The other point is that deploying draft specs on the 'net is a bad idea because eventually the spec goes gold and there are pockets where clients or servers that use the old spec exist which if sizable enough are hard to migrate. Microsoft had this problem with our WD-xsl version of XSL vs. W3C's XSLT.
Dare Obasanjo
Friday, 13 February 2004 15:51:29 (GMT Standard Time, UTC+00:00)
this guy is full of shit. he did n't find this by "reading the blog intermittently"

DeeZ
Friday, 13 February 2004 16:11:48 (GMT Standard Time, UTC+00:00)
"It is quite interesting to see how just one concept [how old is this article?] can lead to some confusion between the end user of a news aggregator and the content publisher. I also suspect that My Yahoo! could be similarly confused by the various issues with escaping content in Atom when processing titles but since I don't have access to a web server I can't test some of my theories."

But that's not a problem with Atom, but rather a conceptual issue with how we decided when an article that is subject to later modification is actually published.
Brian Carnell
Friday, 13 February 2004 16:36:21 (GMT Standard Time, UTC+00:00)
Deez, personal abuse is stupid and pointless.

Dare, I have to agree with Mark here. He's providing accurate data, and just because somebody at Yahoo didn't pay enough attention when sorting the dates, it doesn't mean that Atom is a mess. It just means that an implementor screwed up slightly, resulting in a minor bug.

If some people think that Atom is RSS with a few element type names changed, then they have obviously been paying more attention to the critics of Atom than the actual specification. I doubt many of these people will be implementing Atom support.

> The fact is that most RSS consumers assume that there is one date attached to an item, Atom has three. Different apps will expose this to end users in different ways.

I doubt that. I think that virtually all consumers will choose to display it in the same way as it's been done traditionally on the source websites - ordered by the date it was posted.

There are always teething problems when implementing something for the first time. It doesn't mean that a specification is fundamentally broken.
Jim Dabell
Friday, 13 February 2004 16:46:28 (GMT Standard Time, UTC+00:00)
Jim,
It seems you are the one confusing ATOM as RSS with some tag names changed. There is no notion of "date it was posted" in the ATOM 0.3 syndication spec. There are three dates that can be attached to an entry

1. The date an entry was issued
1. The date an entry was created
3. The date an entry was modified

Technically, none of them is 'date it was posted'. This is just one trivial example of many that show the folks behind the ATOM specs haven't been thinking much about consumers of ATOM feeds while coming up with specs.
Dare Obasanjo
Friday, 13 February 2004 17:40:03 (GMT Standard Time, UTC+00:00)
What is the difference between "posted" and "issued" then? I use the two terms interchangably, and as far as I can tell there is absolutely no difference.

How does it show that "show the folks behind the ATOM specs haven't been thinking much about consumers"?
Jim Dabell
Friday, 13 February 2004 17:45:29 (GMT Standard Time, UTC+00:00)
Shouldn't some of the blame for the confusion be placed at the feet of the cat at Yahoo who whipped out an implementation in half an hour and didn't bother to check his work for logical errors?
Phillip Harrington
Friday, 13 February 2004 17:48:38 (GMT Standard Time, UTC+00:00)
Dare, I thought *you* were one of the folks behind the Atom spec? Don't you participate in the mailing lists and discussions?
Stefan Tilkov
Friday, 13 February 2004 18:23:51 (GMT Standard Time, UTC+00:00)
Wha? Your complaint is that Yahoo! is showing you when items were last updated, not when they were first posted. How is that a problem? It's Yahoo!'s choice -- if you'd like a different setting, file a feature request.

(I think that setting is a good idea -- you can see from the visited status and the title of the post whether you've read the post before, but Yahoo! now also tells you whether the post has been updated since.)
Aaron Swartz
Friday, 13 February 2004 18:44:55 (GMT Standard Time, UTC+00:00)
I do believe that's the fist time I've seen 'Feedster' used as a verb similar to 'Google' !

Oh, and Mark IS full of shit.
casual observer
Friday, 13 February 2004 18:52:56 (GMT Standard Time, UTC+00:00)
I dont' see the problem of the consumer of a feed deciding which order he/she would like to see postings. Isn't this an implementation thing not a specification thing?
Darryl VanDorp
Friday, 13 February 2004 18:58:43 (GMT Standard Time, UTC+00:00)
I think the real point here is that "more is not always better". Just because you can come up with 3 meaningful (sort of) dates about a post doesn't mean that you should include them in a feed. RSS works so great because the vast majority of readers, at least in the beginning, were created by people who were not xml pro's. No dtd's or big hierarchies to worry about. Just a nice simply text file to parse. No need to use xslt or fancy parsers. A couple regex experssions and your done.

Sure this approach probably isn't one I'd use if I was moving financial record data around on wall st, but for getting a couply posting titles and contents on my desktop it's great. There's simply no business case for why I need all this extra data. Title, content, posted date, done.
Ian
Friday, 13 February 2004 19:51:21 (GMT Standard Time, UTC+00:00)
Ian, you don't need "xslt or fancy parsers" to pull three elements out of a feed instead of just one. It's not as complicated as you seem to think.

It's like calling RFC 2822 "a mess" because a mail client you used sorted your email by the time it arrived on your server instead of the time it was sent, and you wanted it the other way around.

Jim Dabell
Friday, 13 February 2004 20:06:33 (GMT Standard Time, UTC+00:00)
Dare, re. the different versions - it's not an easy call, I think it's preferable to have live tests rather than issue Atom 1.0 as a fait accompli only to find it fundamentally broken. That is, assuming it was possible to prevent people from trying the pre-release versions. I don't think the situation is that far from with software releases in general - are you going to continue to support RSS Bandit version 0.0001 indefinitely? All the Atom people can do is make it as plain as possible that 0.3 etc aren't the finished item. In any case, chances are folks that have imlemented already will also be early adopters when 1.0 is ready.

re. your comment: "the current spec is focused too much on content producers and not on content consumers" - you're probably right there, but I still think it's worth trying to get the dates right rather than choosing *today's* lowest common denominator. Incidentally, given Mark's work on the liberal feed parser, he's probably the last person to need a lecture on the subject ;-)

Ian, I hope your horseless carriage is eating well ;-)
Danny
Saturday, 14 February 2004 02:13:02 (GMT Standard Time, UTC+00:00)
Google is premature to try to force developers to support this 0.3 version of atom sice there will be more versions coming and maybe they'll change and again and maybe again and so forth. Looking like future has much more arguing over formats and more boring fights. Why you think there will be less not clear. Seems like confusion is the point of what Google is doing.
Mark Brandes
Saturday, 14 February 2004 07:00:47 (GMT Standard Time, UTC+00:00)
It seems the theme of the day is to complain how much extra work Atom adds for the aggregator creator. Yet Jeremy whipped up his arguably-correct version in half an hour. I realize that everyone may not code at his speed, but it seems to me that the hard part is coding the program to read the first kind of feed. After that, it is all just a matter of modification.

And I think the same sort of market pressure that is causing people to pay attention at Atom in the first place will cause feed providers to move quickly to 1.0 when it comes available.
Michael
Saturday, 14 February 2004 08:51:12 (GMT Standard Time, UTC+00:00)
Michael, Jeremy clarifies in the comments on his blog that he did not actually do the much-talked-about "30 minute implementation" himself. I wonder if we will hear who did.

Some kids watch soaps on TV; I watch the XML dramas. Popcorn!
bystander
Comments are closed.