Shelley Powers writes

For instance, The W3C TAG team -- that's the team that's defining the architecture of the web, not a new wrestling group -- has been talking about defining a new URI scheme just for RSS, as brought up today by Tim Bray. With a new scheme, instead of accessing a feed with:

http://weblog.burningbird.net/index.rdf

You would access the feed as:

feed://www.tbray.org/ongoing/ongoing.rss

I've been trying to avoid blogging about this discussion since I'll be leaving for Philly to attend the XML 2003 conference in a few hours and won't be able to participate in any debate. However since it seems some folks have started blogging about this topic and there  some misconceptions in their posts I've thrown my hat in the ring.

The first thing I want to point is that although Shelley is correct that some discussion about this has happened on the W3C Technical Architecture Group's mailing list they are not proposing a new URI scheme. Tim Bray was simply reporting on current practices in the RSS world that I mentioned in passing on the atom-syntax list.

THE PROBLEM
The problem statement is "How does a user go to a website such as http://news.yahoo.com or http://www.slashdot.org, who'd like to subscribe to information from these sites in a client-side news aggregator do so in a quick and painless manner?". The current process is to click on an icon (most likely an orange button with the white letters 'XML' on it) that represents an RSS feed, copy the URL from the browser address bar, fire up your RSS client and click on the subscribe dialog (if it has one).

This is lot of steps and many attempts have been made to collapse this into one step (click link and the right dialog pops up). 

PREVIOUS SOLUTIONS TO THE PROBLEM
The oldest one I am aware of was pioneered by Dave Winer and involved client side aggregators listening on a port on the local machine and a hyperlink on the website linking to a URL of the form http://127.0.0.1:5335/system/pages/subscriptions. This technique is used by every Radio Userland weblog and is even used by dasBlog which is my blogging tool of choice as is evidenced by clicking on the icon with a picture of a coffee mug and the letters "XML" on it at bottom of my weblog.

There are two problems with this approach. The first is the security issue brought on by the fact that you have a news aggregator listening on a socket on your local machine which could lead to hack attempts if a security exploit is found on in your news aggregator of choice, however this can be mitigated by firewalls and so far thus hasn't been a problem. The second is that if one has multiple aggregators installed there is contention for which one should listen on that port. For this reason different aggregators listen on different local ports; Radio listens on port 5335, AmphetaDesk listens on port 8888, Awasu listens on port 2604, nntp//rss listens on port 7810 and so on.

An alternative solution was chosen by various other aggregator authors in which hyperlinks pointed to the URLs of RSS feeds with the crucial distinction that the http:// part of the URL was substituted with a custom URI scheme. Since most modern browser have a mechanism for handing off unknown URI schemes to other client applications this also allows "one-click feed subscription".  Here also there is variance amongst news aggregators;  Vox Lite, RSS Bandit & SharpReader support the feed:// URI scheme, WinRSS supports the rss:// URI scheme while NewsMonster supports the news-monster:// scheme.

With all this varying approaches, it means that any website that wants to provide a link that allows one click subscription to an RSS feed needs to support almost a dozen different techniques and thus create a dozen different hyperlinks on their site. This isn't an exaggeration, this is exactly what Feedster when one wants to subscribe to the results of a search. If memory serves correcly, Feedster uses the QuickSub javascript module to present these dozen links in a drop down list.

THE FURORE
The recent debate on both the atom-syntax and the www-tag mailing lists focuses on the feed:// URI proposal and it's lack of adherence to guidelines set in the current draft of Architecture of the World Wide Web document being produced by the W3C Technical Architecure Group. This document is an attempt to document the architecture of the World Wide Web ex post facto.

Specifically the debate hinges on the guideline that states

Authors of specifications SHOULD NOT introduce a new URI scheme when an existing scheme provides the desired properties of identifiers and their relation to resources.
...
If the motivation behind registering a new scheme is to allow an agent to launch a particular application when retrieving a representation, such dispatching can be accomplished at lower expense by registering a new Internet Media Type instead. Deployed software is more likely to handle the introduction of a new media type than the introduction of a new URI scheme.

The use of unregistered URI schemes is discouraged for a number of reasons:

  • There is no generally accepted way to locate the scheme specification.
  • Someone else may be using the scheme for other purposes.
  • One should not expect that general-purpose software will do anything useful with URIs of this scheme; the network effect is lost.

The above excerpt assumes that web browsers on the World Wide Web are more likely to know how to deal with unknown Internet Media Types than unknown URI schemes which is in fact the case. For example, Internet Explorer  will fallback to using the file extension of the file if it doesn't know how to deal with the provided MIME type (see  MIME Type Detection in Internet Explorer for more details). However there are several problems with using MIME types for one click feed subscription that do not exist in the previously highlighted approaches.

Greg Reinacker detailed them in hist post RSS and MIME types a few months ago.

Problem 1: [severity: deal-breaker] In order to serve up a file with a specific MIME type, you need to make some changes in your web server configuration. There are a LOT of people out there (shared hosting, anyone?) who don't have this capability. We have to cater to the masses, people - we're trying to drive adoption of this technology.

Problem 1a: [severity: annoyance] There are even more people who wouldn't know a MIME type from a hole in the head. If Joe user figures out that he can build a XML file with notepad that contains his RSS data (and it's being done more often than you think), and upload it to his web site, you'd think that'd be enough. Sorry, Joe, you need to change the MIME type too. The what?

Problem 2: [severity: deal-breaker] If you register a handler for a MIME type, the handler gets the contents of the file, rather than the URL. This is great if you're a media player or whatever. However, with RSS, the client tool needs the URL of the RSS file, not the actual contents of the RSS file. Well, it needs the contents too, but it needs the URL so it can poll the file for changes later. This means that the file that's actually registered with a new MIME type would have to be some kind of intermediate file, a "discovery" file if you will. So now, not only would Joe user have to learn about MIME types, but he'd have to create another discovery file as well.

Many people in the MIME type camp have pointed out that problem two can be circumvented by having the feed file contain it's location. Although this seems a tad redudundant and may be prone to breakage if the website is reorganized it probably should work for the most part. However there is at least one other problem with using MIME types which people have glossed over. 

Problem 3:  If clicking on a link to an RSS feed in your browser always invokes your aggregator's feed subscription dialog then this means you can't view an RSS feed in your browser if you have a client aggregator installed and may not be able to view even if you don't because your browser of choice may not know how to handle the MIME type if it isn't something like text/xml.

At least one person, Tim Bray, doesn't see this as a big deal and in fact stated, "why not? Every time I click on a link to a PDF doc I get a PDF reader. Every time I click on a .mov file I get Quicktime. Seems like a sensible default behavior".

THE BOTTOM LINE
Using MIME types to solve the one click subscription problem is more diffficult for weblog tools to implement than the other two approaches favored by news aggregators and requires changing web server configurations as well while the other approaches do not. Although the architecure astronauts will rail against the URI scheme based approach it is unlikely that anyone who looks dispassionately at all three approaches will choose to use MIME types to solve this problem. 

Of course, since one of the main forces behind the ATOM movement has stated that MIME types will be the mechanism used for performing one click subscription to ATOM feeds this just seems like one more reason for me to be skeptical about the benefits of adopting the ATOM syndication format.


 

Categories: XML
Tracked by:
http://www.xt1.org/index.php?p=31&c=1 [Pingback]
http://www.25hoursaday.com/weblog/PermaLink.aspx?guid=7ae505bc-fba5-43a9-b73e-13... [Pingback]
http://www.25hoursaday.com/weblog/PermaLink.aspx?guid=c5e678a4-7e74-4f13-bb68-57... [Pingback]
http://www.25hoursaday.com/weblog/PermaLink.aspx?guid=2a056f3b-de38-4751-85e4-be... [Pingback]
http://www.25hoursaday.com/weblog/PermaLink.aspx?guid=6deaa5e1-46b5-4e19-a9b9-57... [Pingback]
http://silauma.info/washington/sitemap1.html [Pingback]
http://jawf5j3.net/sublets/sitemap1.html [Pingback]
http://dugablog.com/sitemap1.html [Pingback]
http://otjjblj.net/01/index.html [Pingback]
http://weujmru.net/pets/sitemap1.html [Pingback]
http://weujmru.net/sublets/sitemap1.html [Pingback]
http://ukpuuq8.net/musicians/sitemap1.html [Pingback]
http://restablog.com/garden/sitemap1.html [Pingback]
http://exktlmy.net/linux/index.html [Pingback]
http://lx2rnws.net/childcare/sitemap1.html [Pingback]
http://harlleyd.extra.hu [Pingback]
http://imprus.iquebec.com [Pingback]
http://tkru7ln.net/sitemap1.html [Pingback]
http://fl7vaa4.net/internet/sitemap1.html [Pingback]
http://restablog.dreamhosters.com/lessons/sitemap1.html [Pingback]
http://tulanka.readyhosting.com/travel/sitemap1.php [Pingback]
http://host239.hostmonster.com/~blogford/sitemap3.html [Pingback]
http://host239.hostmonster.com/~blogford/sitemap2.html [Pingback]
http://gator413.hostgator.com/~digital/babies/sitemap1.html [Pingback]
http://gator413.hostgator.com/~digital/gift/sitemap1.html [Pingback]
http://da7fcil.net/15/index.html [Pingback]
http://fskh6vo.net/sitemap1.html [Pingback]
http://gh9kwkn.net/myspace/sitemap1.php [Pingback]
http://xu6r4om.net/sitemap1.html [Pingback]
http://qyaq5qm.net/mortgage/sitemap1.php [Pingback]
http://qyaq5qm.net/truck/sitemap1.php [Pingback]
http://d579737.u108.floridaserver.com/sitemap2.html [Pingback]
http://d579737.u108.floridaserver.com/sitemap1.html [Pingback]
http://gator442.hostgator.com/~hockteam/alaska/sitemap1.html [Pingback]