Trackbacks vs. Referrers

John Gruber's critique of Trackback makes some good points. And I agree with many of them. But I have to disagree with his central argument that you can use referrer data instead. This argument completely misses what I feel is the best use of Trackbacks: category-specific aggregation. Take Blogpopuli, for instance, which uses Trackback to gather posts about weblogs. Each of the site authors who send Trackbacks to Blogpopuli would have absolutely no reason to link (create a referrer) to Blogpopuli every time they posted about weblogs—it just wouldn't make sense to their post. By sending a Trackback behind the scenes, they let a subsection of the world know the post is about weblogs without changing the substance of their post. It's a subtle distinction, perhaps, but I think an important one. Using Trackbacks, we suddenly have a new service that simply points to distinct posts of a given category. Extreme-niche, category-specific aggregators can be set up using Trackback that just aren't possible with referrers or (non-categorized) RSS. And an open posting form means even those who aren't using Movable Type can contribute.

Even with its current widespread use of Site A linking to Site B, referrers don't measure up. For a referrer connection to be made, someone has to click on the link. That may seem trivial, but it means less popular sites could be cut out of the loop. Referrers track what a site's audience is doing, not what the site's author intended. To make matters worse, the referrer will not point to the specific post that is relevant. For example, most people will click a link from this weblog's main page, and create the referrer: As posts slide off the front page, referrer data will quickly become meaningless. Trackbacks use permalinks, which means the link will work well into the future. (Not to mention the problem most referrer-tracking scripts have distinguishing between and're the same site!)

There are also some cases were it's impossible for referrers to work. Check out the way Matt Haughey is using Trackbacks to keep a list of songs he's listening to in Winamp. Because Winamp isn't a website, referrers can't exist. Extending weblogs into other devices requires a mechanism for transporting Web micro-content; Trackback could be that mechanism.

Sure, there are some problems with Trackback. And I think its current primary use of Site A notifying Site B of a relevant post isn't really its killer use. Topic-specific weblog post-aggregation is where Trackback could be very powerful—and relying on referrers isn't going to work to make that happen.


It's interesting that you prefer category aggregation TrackBacks; That was the original impetus behind the creation of the protocol and the per-post (remote comments/referrers) use only arose later as one implementation.
I had several conversations with Ben and Mena about this very subject of category-specific post-aggregation before Trackback was implemented, so I'd like to think I helped influence its implementation. But once a part of MT, I was surprised at how it was being used...basically as a better referrer. I think hooking some known category aggregators (like Blogpopuli) into any weblog tool interface would help encourage Trackback's category-style uses.
I've been very happy that the Seattle Webloggers portal-thingy implemented a category aggregation trackback form. Now Seattle weblogs can have their specifically-Seattle-related posts listed!
If you look at Gruber's page again you can already see some of the problems with referrer tracking:

The former is redundant and without context, and the latter is a password-protected newsreader.

And I particularly like these:

I also notice that Technorati doesn't distinguish between a static, run-of-the-mill link (like a blogroll entry) and a relevant link like a weblog post or article, so I don't see how it's any better, other than the fact that it proactively looks for links.
All the static-rendering problems mentioned in Gruber's otherwise fine article can also be applied to comments on weblogs.

Comments are seen as absolutely fundamental to almost all weblogs and it's what turns weblogs into communities, but you can't have comments in a Blogger blog natively. You can't with Radio either, but you can with centralized servers provided by many third parties (and Userland provides its own to Radio users). I'm kind of surprised that central trackback services haven't popped up much like comment services have, but I guess it has less of an appeal and it has the "it takes two to tango" problem.

I agree with Gruber that Technorati does a bunch of great stuff, on my personal site I don't even look at my stats anymore, I just paid the $5 to Sifry to have a nightly email with recent mentions sent to me, but I have to say that almost half of all the links to any of my sites in technorati aren't related to blog posts, and are not nearly as useful. If someone adds you to their blogroll, it shows up in Technorati. If someone added you two years ago and just made their first ping (which I have to say is the weakness of Technorati, anyone not pinging doesn't exist in that system), then you also get notified. I much prefer posts on others sites that mention something specific that I have said, and for that, trackback kicks ass.

Also on my personal site I have my own personal trackback portal of stuff I write on other sites. I'm going to ping it for this comment, which you should find here, on the lower right under "posted elsewhere".
I just found a hosted trackback system that works much like a hosted comments system.
"You can't with Radio either, but you can with centralized servers provided by many third parties (and Userland provides its own to Radio users)."

What a convoluted bit of logic.

Radio has comments. What else do you want from Radio. Geez Louise.
One of the original article's critiques of Trackback is that it requires a CGI script to generate the list of Trackbacks. So in that way it's "not native" to a static HTML site. And Matt's point is that services like Radio blend static and generated, on-the-fly content all the time to enable comments. It wasn't a critique of Radio...just a technical parsing of how the systems work to make his point. (Local and remote servers working together.)
Dave, my "convuluted logic" stems from this passage in Gruber's piece that describes how impossible it would be to add trackbacks to radio:

"Radio UserLand, for example, runs on your desktop. It's actually a web server, but it runs on your machine, and then uses FTP to push your weblog files to the web as static HTML files. This means to receive TrackBack pings, your desktop computer would need to be publicly accessible over the Internet, 24 hours a day. Not only is this unlikely, it's ill-advised. If you're on dial-up, you're not going to be connected 24 hours a day. And even if you have a broadband connection, you probably don't have a static IP address, and there's a good chance you're behind a firewall of some sort."

My point is that the same could be said for comments on Radio blogs, but they work. My use of the modifier "native" is due to Radio not having comments by default when the last version was released in January of 2002. Radio Community Servers came out three months later:

Again, my point is that Gruber's criticism of Radio and Trackback is wrong, as evidenced by the adoption of comments in Radio. Trackback could be handled in an identical way.
I am still not seeing a great advantage for trackback. Having read Gruber's comments I am somewhat intrigued by the continual positive comments surrounding trackback and the very limited use. It seems that the Seattle Weblog site could pull an RSS feed for member's sites that use a Seattle category. This is done at ia/ by many and aggregated into a side module. I have run across many other sites that do the same thing and nearly all weblog tools have RSS builders now and most have categories.

Every time I go to build trackback capabilities into my own tool, I realize there is already another method for performing the same functionality. I have hand coded a couple trackback pings thinking it was the thing to do and I really needed to code one for myself, but I was underwhelmed with the result of trackback.

Granted few read their access logs, but a referrer tracking for each permanent entry could do something similar. Yes, there are varied entries from the same page, but one line in a string could take care of that.

Just my perspective.
But even categorized RSS is clumsy when compared with the elegance of Trackback. How many RSS feeds currently have any sort of categorization for their posts? Next to none, in my experience. And how many RSS feeds would need to be monitored to create a service like BlogPopuli or Seablogs? Hundreds? Thousands? The way it works now, the weblog authors can simply notify the site when they create a relevant post. The bandwidth wasted on brute-force aggregation by scraping every potentially relevant RSS feed would be insane. Why have any similar service scrape hundreds of feeds looking for the chance mention of the category it's interested in, when a simple ping would do the job?

And a referrer tracking for each permanent entry would still not link to the permalink for the referring site. (Try parsing that sentence three times fast. ;)

I agree with you that there are few implementations of Trackback that illustrate its advantages. But just because the perfect application for Trackback hasn't come along yet doesn't mean it doesn't solve a unique problem. I really believe simple notifications like Trackback pings would make aggregating similar posts much easier than scraping hundreds of RSS feeds.
Point well taken with the syndication aggregator. With RSS there are some of us that have some categorical RSS feeds established in addition to our full blog RSS. This would allow more than just the folks getting the trackback notification to have access to this feed with out any additional work on the part of the owner. Many have created such RSS feeds to be sucked in by ia/ ( ) and the current aggregation is posted in the default right sidebar. Should another site popup for Seattle weblogger the content creators would not need to set a new trackback ping, but the new site could just pull in the Seattle information. Part of what makes this doable is the ability to assign more than one category to each post (in a keyword or heirarchical approach), not not all blog tools offer. (I have readers of my site that only read one or four categories and skip the rest, as to them it is not of use.)

The referrer tracking only needs to have one click-through from the permalink page to the permalinked article to register properly. This click-through can be done by hand or automated, similar to trackback. By default I click-through all my links after posting to verify they work and go to where I thought they would go.

I keep thinking trackback could solve the external comment tracking issues. When I post a comment, say here in this thread, I would like to know when others have continued on in the discussion. I have the capability in my tool to do this via e-mail, but have not made it live and have archieved the code. Sometimes I can not remember where I left a comment to even go back and manually go see if there have been new additions to the discussion. Trackback seems like it would be a good solution for this even though some folks have set up RSS feeds for individual comment pages, which works to some degree.
Ahh, I see what you're saying now with the referrer coming from the archive page for a post...though that feels kind of hacky to me. (Sort of like news aggregators that set the referrer when they grab a feed.)

I think this whole thing boils down to your point of view: if you're a weblog author you don't want to go through the extra work of pinging aggregators. If you're an aggregator-builder, you don't want to go through the extra work of scanning unnecessary feeds to find that 2% of the content that is relevant.

And I see what you're saying about pre-categorized RSS...thanks for the example. I think RSS is great when all of the content is relevant to the syndicator. And Seablog with an RSS feed means another seablogs site has less work to do. But when only pieces of the content (like weblog posts) are important to an aggregator, Trackback works well for pointing at specific parts. Say someone sets up a site for a specific neighborhood in Seattle, they only want posts relevant to that part of town. The whole seablogs RSS file is a good source, but can't be used as a whole...that's why setting up a separate ping URL would be good.
i agree that RSS aggregation is a better syndication solution when you know all the sources you are trying to aggregate. One of Radio's strengths, for example, is the ability to autopost selected RSS feeds to teh Radio blog.

but trackback ping aggregation seems to make more sense when you are opening it up to anyone who wants to contribute without requiring some avanced listing of contributing feeds.

Search Results

No emoji found