Time to get over “time to get over comments”

Over and over again, you hear it:

Personally I think this whole mess is just gonna go away. Everyone of you should have your own blog. If I say something idiotic, you should blog it, link to me. I watch my Technorati page and Feedster and Pubsub too. That’s a much better way to comment on something I write.

I keep thinking that it will go away, too. I keep thinking that people will actually think it through, and realize how unlikely and unsatisfactory that solution is, and we’ll stop seeing people calling for an end to weblog comments.

My first objection is to the arrogance of that position: “of course, if you want to comment on my entry, all your readers would be interested in both your comment and my entry.” Sorry, no. In a linkdump post yesterday, Scoble claimed that Browse Happy is “The Mozilla guys” when as you already know from being pointed to it from a dozen directions it’s actually The Web Standards Project guys and gals. Did you really need another link to Browse Happy, which you already knew about (and knew that you weren’t its target market), just so I could tell both Scoble and you that it was done by the people linked at the bottom of the page, not by one of the browsers it recommends?

My second objection is to, well, the arrogance of that position. Robert says that he watches lists of his incoming links. So what? If my goal was to be sure that he knew Browse Happy was WaSP not Moz, I would have emailed him. What I wanted was to be sure that both he and his readers knew, whether or not he felt like going back to correct his post. The totality of a post and its comments are where we all arrive at as true a picture as we have time and energy to paint. The “I post, you post, I see you linking” model sounds more like you take a snapshot of what you see and show it to us, we take snapshots and show them to you, and maybe the person next to us sees our snapshot while we show it to you, but nobody sees enough to put together a complete picture.

Of course, Scoble could continually edit his posts, adding in links to everyone he sees linking to him, along with at least an abstract of what they said, if not the full content of the post. I’ve actually seen that done, and I imagine it was nearly as annoying for the person doing it as it was to read. If you posted about something that didn’t happen to interest me, the last thing I want is to see that post updated a dozen times, or a dozen new posts saying “Foo linked to that post you didn’t care about, and said bar.”

A post that does interest you is at least as bad. You want to read what other people had to say about it, so you either follow all the followup links the original blogger posted, or use Feedster and PubSub and Technorati and Bloglines Citations to get four sets of only partly overlapping links to the post. Then, because each person responding to the original had to give their readers some context for what they were responding to, you first have to read a dozen recaps of the original post. How could we avoid that? D’oh! By putting all the responses where the authors know everyone reading them has already read the original, as comments on the original post.

Oh, the “this whole mess”? Robert’s using Javascript powered comments remotely hosted on one of Dave Winer’s servers, done by making a remote Javascript call for the count for each comment thread separately, so when the comment server’s slow or inaccessible it makes his page loads incredibly slow, as each call has to time out before the page load can continue. Um. That’s a problem that all the rest of us solved several years ago: notice how every other remote comment system loads the counts in one single Javascript file? If you remember far enough back, Robert, the first time you noticed my existence was when I was telling you how to make that more robust. Shame you went to something far, far less robust ;)

Oops. Teach me to think I understand what someone’s saying, and they understand the problem, without checking. Jake says I got it wrong, the comments were doing the reasonable thing and loading a single script, but a single script of counts for every Scoblepost is apparently an enormous wad of data to pull out of a database and send out with every pageload.


Comment by michel v #
2004-08-23 04:56:43

I feel so compelled to blog ”I agree with Phil.”

I’ve always found the ”get a blog” attitude stupid, mostly because many comments are made by non-bloggers who really have no special interest in maintaining a blog just to comment on other people’s posts.
And then there’s that whole overlapping sets of links problem: if I am linked to by a personal page on Geocities that’s not a blog and not crawled by any blog-crawler, I’m the only one who’ll know about the comment. Yeah for transparence and free debate.

Partly off-topic:
However, I have a reverse idea: listing comments you make on other blogs, on yours. Think of it as ”I didn’t post much today, but I put some words over there, that didn’t warrant an actual blog entry.”
That could be done pretty much automatically, or manually with a bookmarklet once you’re sure you’re going to hit the submit button.

Comment by Phil Ringnalda #
2004-08-23 08:39:09

To a certain extent, blogging my comments was what made me start a linklog. Partly for exactly what you said: ”crap, I used up all my writing time and effort on other people’s blogs today” (er, this week/month), and partly just because lots of times when I leave I comment I want to remember to check back later, and this is where I put URLs that I need to remember. Of course, like all my good intentions, I don’t actually do it as often as I should.

Comment by Anne #
2004-08-25 12:11:57

I just noticed I linklog my short comments instead of commenting and linking to them. The problem is that my linklog doesn’t have pingback enabled. It probably should.

Comment by Paul Watson #
2004-08-23 08:45:09

”Get a blog” isn’t a great attitude but the current state of blog commenting isn’t terribly workable either. One may visit ten posts, leave behind ten comments and then wander off and do some actual work. A few hours later you wonder if anyone has responded to your comments. What then? You have to have bookmarked those posts or go through your history finding those ten links. Getting email notifications is a bit much too, a popular post you comment on which does email notifications could flood your inbox. Also most sites don’t include comments in their RSS for a given post (or if they do and each post has it’s own RSS then how often do you go to the hassle of temporarily subscribing to that post?)

How about using TrackBacks more effectively and in tandem setting up a GotNoBlogButWantToComment.com site.

If someone has a blog and wants to comment, they post on their blog (or a side-blog if they don’t want to clutter their main-stream) and send a TrackBack ping to the post. Problem solved for them. They can both comment and get notifications (through TrackBacks) of replies to their comment.

If that someone does not have a blog then there is a normal enough looking Comment form on the post. But the form isn’t part of the original post’s website but rather of GotNoBlogButWantToComment.com. It asks for the normal Comment fields but also a password (can be TypeKeyed or be a ”throw away” account and have no password). On submit GotNoBlogButWantToComment.com checks if the user has a blog already, if so then it logs them in and carries on. If no blog is present one is transparently setup and the Comment/Post made. It then automatically sends a TrackBack to the commented-on post (which lists TrackBacks like comments are now).

A nice side-benefit of this is that should you want to comment on a comment you have a permalink (it’s a blog post afterall) and it has a TrackBack list so you get some handy threading. Also one post can TrackBack multiple posts so you get a web of commenting, not just hierarchical threading.

It is all then very easy to:
– See all comments you have ever made on any site
– Replies to those comments
– Edit your comment should you need to
– Avoid getting spammed and not having to leave behind a fake email address on those comment systems that require one
– and GotNoBlogButWantToComment.com can easily check the original post you are commenting on for new comments and notify you

Just an idea :)

And now I wander off to get some work done and in a few days will have completely forgotten which site I just commented on and so won’t be able to come back and check if people are calling me crazy yet or not (so go call me crazy on my blog if you wish).

Comment by Roger Benningfield #
2004-08-23 10:52:00

Paul: I knew you posted this comment because Phil’s wonderful feed uses the wfw:commentRss extension. I don’t need to manually subscribe to anything… Newzcrawler automatically fetches Phil’s comment feeds and lets me know whenever someone posts something new.

Solutions are out there, and they work… it’s just a matter of lobbying your favorite bloggers to do the world a favor and add one measly little element to their RSS templates.

Comment by Roger Benningfield #
2004-08-23 11:37:05

Paul: And while I’m thinking about it, most of what you describe as GotNoBlog* already exists.

Go create a JournURL account, login to the Support community, post a reply to Phil in the Sandbox category, and put the URL of this entry in the ”annotate” field. JournURL will send a Trackback from that post to Phil’s blog. Phil can then reply directly to your post by logging in himself, or he can send a ping from his blog to that post’s permalink.

Or we could decentralize the process a bit. I could subscribe the Sandbox category to Phil’s feed, and all of his posts would automatically show up as individual threads within the forum. Phil’s blog could then auto-subscribe to those threads’ respective feeds and display any new posts alongside his native comments.

Of course, GotNoBlog would be misleading at that point, since any of your posts can also be weblog entries in the blog that was created for you the first time you signed in.

(Thanks to Phil’s forced preview, I just reread the above. It’s almost impressive, the way I can take an incredibly simple, two- or three-click process and come up with such an arcane-sounding description. There should be an award for that sort of thing… I’d win every year.)

Comment by Paul Watson #
2004-08-24 01:58:54

Thanks for the heads up on the RSS comment tag, never seen it before :)

As for the rest, you are right in that it’s pretty much all there but the main problem is the ”transparent” registering of a comment blog. If the commentor has to go through a Blogger or Live Journal setup screen just to comment, they won’t. And I doubt Blogger or LJ will open up their registration process to allow for this as they may fear a rash of spam registrations.

What you say all sounds pretty good to me though.

Comment by ben #
2004-08-23 11:36:18

[What follows is a near-dupe of the comment left at Scoble’s place.]

More background on Browse Happy has been published at http://www.webstandards.org/buzz/archive/2004_08.html#a000413 and http://www.webstandards.org/act/campaign/happy/ – it’s just about the shifting sands of technology, really. In ’99 and ’00 IE5 was the bee’s knees because it was doing a comparatively great job… but now the Microsoft browser is the one lagging behind, period-end-of-sentence.

Comment by Jake Savin #
2004-08-26 15:50:06


The fact of the matter is (or was until Scoble turned his counters off) that his comment counts were being loaded as a single JavaScript file. The problem in this case is not that there was a separate JavaScript for each comment count, but rather that the server was serving counts for all of his posts, not just the ones on the page being loaded. A quick View Source would have revealed that.

In Scoble’s case, the sheer number of posts he made, and the frequency with which people commented is what brought his comment counter up against this scaling issue.

This is one of the things that UserLand is working on solving for Radio in an upcoming maintenance release. I don’t have a specific date yet, but it will be soon.

Comment by Phil Ringnalda #
2004-08-26 16:19:14

Ah, sorry for the slur, Jake. I just misinterpreted

Well, Radio Userland displays comments using a JavaScript that calls into the Manila server that’s hosting my comments. Problem is that the browser needs to communicate with the server for each blog post to get the numbers of comments for each post and the page wouldn’t display until it had all of them. So, when the comment server started going screwy, it kept my whole page from loading.

as meaning ”communicate with the server once per post in the page” rather than ”communicate with the server about each and every post.”
Somehow, I’ve fallen so far out of the Radio community that I couldn’t think of anyone else using Radio comments (which wouldn’t necessarily be done the way scoblecomments were done) to check whether he really meant what he seemed to be saying. And, of course, I’m a vicious bastard, so I assumed the worst. Sorry about that.

As I remember it, when Hossein hit that same problem with YACCS, he started only returning counts for (err, memory fails) the most recent
100 posts? Posts from the last two weeks? Something like that, with the count function returning something like [archived] for any post where it didn’t have a count. Of course, there was great wailing and gnashing of teeth, but as anyone with archives stretching back into the dusty past knows, if it ain’t on your main page or in your feed, it ain’t getting comments. Well, except for spam.

Comment by Jake Savin #
2004-08-26 23:47:28

Well the nice thing is that when the revisions to UserLand’s comments server are released, we won’t have to worry too much about when comments or the page that links to them were created — the server, with a tiny bit of cooperation with the client, should be able to send the counts for only, and exactly the posts on the page being viewed.

This in combination with a little modernization of our HTTP header treatment, and performance should improve by a good bit. I’d venture to guess that Scoble could turn his counters back on without becoming ungodly slow again, but time and testing will show whether I’m correct about this or not.

Thanks by the way for updating your post, Phil.

Comment by Marcus #
2004-08-26 16:58:26

You’re sending all the comment counts… ever? There’s a blast from the past. I had a wee check and it seems I had to start limiting counts in BlogBack back in March 2002. (I chose to return the counts for the last 30 posts, which is usually ~1KB.)

Comment by Jake Savin #
2004-08-26 23:54:29

Yeah, it’s a mess, which is I guess no surprise to anyone at this point. ;-)

The funny thing is that for our comments server (actually running in Manila), how many comments or posts there are for a given site aren’t the primary limiting factors. Rather in this case, the sheer size of the JavaScript that was being returned when Scoble’s pages were requested was what caused the extreme slowness.

The way our database is structured, finding the counts in the database is relatively quick. It was the actual transmission of the JavaScript content that took up most of the computing power. This means that we don’t have to limit to recent counts, but rather just to the relevant counts.

In other words, it should be transparent to the person browsing the site. You might want to consider doing something similar with your comment counters, if your database permits it within reason.

Comment by Phil Ringnalda #
2004-08-27 00:20:45

Funny thing is, I was fully prepared to say that while you have the advantage of having control over the blogging engine, so you can just
generate a script tag that includes the item numbers, supporting Blogger (which is what most remote comment hosts mostly do) isn’t that easy, since you don’t have any way to do it. Then for once I actually caught myself, and checked on whether or not what was true in 2002 was still true. view source (not that the link gives you a choice) and look at the script element right after the <head> tag. You can now have multiple <Blogger></Blogger> sections!

Comment by Marcus #
2004-08-27 04:37:50

Cool. If only they’d done that years ago (I never could understand why the Blogger tag had to be limited that way).

Comment by Phil Ringnalda #
2004-08-27 11:17:59

Be darned. Speaking of things changing, apparently IE6XPSP2 (at least) disables (in a very unhelpful way) the view-source: pseudo-protocol. Bummer. Have to do it yourself, I guess, if that leaves you with an error page suggesting that you ”open the res://C:WINDOWSsystem32shdoclc.dll/syntax.htm#view-source:http:// home page”.

Comment by Marcus #
2004-08-27 04:34:52

Aye, there’d be no problem with the database; I actually had to improve the way the comments counts were being stored before I even worked out that the file I was sending was way too large.

I really wanted the context idea to work at the time, but it was a no-go… especially because my users aren’t using the one system. Depending on what’s being used, it’s hard enough to get a handle on a ID for each post, let alone for all the posts visible on one page.

Comment by Marcus #
2004-08-27 04:47:24

Actually, having a look at Phil’s Blogger example, maybe I shouldn’t discount the idea so readily. I had been thinking that I wouldn’t be able to add this without maintaining two vastly different versions of code — Blogger / non-Blogger — but it’s probably not as awkward as all that. All I’d have to do is add an extra variable for them and, if that’s being used, alter the SELECT a wee bit. Hmm…

Comment by Marcus #
2004-08-27 11:01:31


Name (required)
E-mail (required - never shown publicly)
Your Comment (smaller size | larger size)
You may use <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <del datetime="" cite=""> <dd> <dl> <dt> <em> <i> <ins datetime="" cite=""> <kbd> <li> <ol> <p> <pre> <q cite=""> <samp> <strong> <sub> <sup> <ul> in your comment.