On 1/30/06, Jens Alfke <email@hidden> wrote:
> Isn't "application/rss+xml" a de-facto standard?
Not really, as a very large proportion of people use text/xml or variants.
But in any case, a MIME
> type wouldn't suffice. A feed is more than a static document like a JPG: its
> value comes from the fact that it changes over time in a meaningful way, and
> the value of a news-reader is that it tracks those changes.
>
> Assuming a standard MIME type, and assuming news reader apps that register
> that type with the OS and the web browser, the user experience would be
> 1. I click on that silly orange "XML" button
> 2. This "index.rss" turd-file appears in my downloads folder
> 3. A news-reader launches and displays the articles in the feed.
> 4. The feed never updates in the news-reader because it has no idea where to
> refresh it from.
>
> Step 2 is annoying, but 4 is a deal-breaker.
Both of these are implementation details of browsers. As you note,
Atom has a <link rel="self"... primarily as a workaround for this
case. But the long-term solution is not to encourage something new
that works contrary to current best practice, but fix the underlying
problems. Most people that have looked at the problem in this light
have seen autodiscovery as the best way forward. Get rid of the silly
orange button altogether.
Barring some kind of
> nonstandard out-of-band communication (like the browser adding the URL of
> the feed to the file's metadata) the news reader cannot subscribe to the
> feed, because subscribing requires knowing the feed's URL, not just its
> contents.*
>
> My point is that a news feed is not just a document. It is a dynamic
> resource that requires a protocol to make use of it, much like a POP
> account. Therefore I think that identifying it via a URL scheme is
> appropriate.
There's the informal hourly polling protocol, but that uses HTTP. The
representation of the feed resource changes over time, sure. But there
is no such thing as a URL scheme, URIs are identifiers, orthogonal to
the protocol.
As Tim Berners-Lee put it:
[[
My feeling is that feed: is basically harmful in that it confuses the
identity of the object with the way the user should treat it. "feed:"
and "webcal:" are harmful in the same way, and "http:" should be used.
]]
http://lists.w3.org/Archives/Public/www-tag/2004Feb/0004
> Now, i think some of the confusion here stems from the fact that, for better
> or worse, the current protocol for subscribing to a news feed involves
> nothing more than repeatedly fetching it as a normal HTTP resource. But that
> doesn't mean a feed should be treated as any other HTTP resource.
If the feed is being delivered over HTTP, then it is just like any
other HTTP resource.
> *[Atom does work around this with <link rel="self">, but this doesn't change
> my point that the location of a dynamic resource is more important than its
> contents.]
Browsers could behave in a more useful fashion.
Cheers,
Danny.
--
http://dannyayers.com
_______________________________________________
Do not post admin requests to the list. They will be ignored.
syndication-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/syndication-dev/email@hidden
This email sent to email@hidden
