Chad Perrin: SOB

4 June 2008

What’s wrong with reddit?

Filed under: Geek — apotheon @ 12:53

I wonder if reddit had problems like this back when it was written in Lisp instead of Python:


Notice how the number of points indicated on the headline (1) differs from that indicated in the box on the right (4 – 1 = 3).

I’ve seen a number of weird little problems like that at reddit over time. Another one I’ve seen:


Those are just two issues I’ve run across in the last couple days, of course. There have been other inconsistencies and errors. Despite all this, however, things aren’t as bad as they are at Digg. At least at reddit pages consistently load in under ten seconds, the submissions process isn’t user-hostile, and I can actually find everything I need within a reasonable amount of time. Hell, I still haven’t found a “bookmarklet” for Digg. All of that is completely aside from the fact that Digg’s content tends to be considerably more low-brow, vapid, pointless — to say nothing of the much higher quality of discussion at reddit. Yeah, there are trolls and morons in reddit discussion, too, but I’m regularly surprised at the intelligent humor and knowledgeable basis for some of the arguments.

. . . but the reddit errors annoy the hell out of me, sometimes, and they seem to be growing more common. Maybe it’s just a subjective illusion. Maybe not, though — and if not, I wonder why the problems are getting worse. Probably some kind of issue with replication of data across servers, I guess.


I find the disagreement between the languages I mentioned and the language mentioned in the ad I found at the top of this page amusing too:

ad shows C++, content shows Lisp and Python


  1. I believe that since you’re sorting by “hot” that it takes into account some sort of popularity scoring in addition to the up/down votes. I think “top” is what you want to just count the votes. I think.

    I like being able to see the amount of up/down votes now, as well seeing the most controversial posts. Good, interesting changes.

    Comment by Brandon — 5 June 2008 @ 05:49

  2. Yes, because we all know Python is doing the math incorrect.

    Seriously…are you that dumb? Ever hear of caching? I’m positive more people hit the front page then the comments page, so they cache that and retrieve the comments page count near real time.

    But I’m sure if they wrote in in RoR, it would have worked correct out of the box. And if they did it in Lisp, reddit would have gotten PhD’s for everyone who views the site…and in PHP it would have failed b/c no great site works in PHP (wait…facebook, digg, etc).

    Learn concepts about scaling web applications before spewing your spray about your favorite language.

    Comment by bill — 5 June 2008 @ 06:35

  3. Brandon:

    Are you saying you think reddit scores things differently depending on how you’re sorting things and what page you’re viewing?


    Speaking of dumb — I think you completely misunderstood most of what I said. Perhaps you could try reading again before setting up, and attacking, these absurd straw men. Good job making an ass of yourself.

    addendum: I decided to get into just a tiny bit more depth, in response to bill.

    I’m positive more people hit the front page then the comments page, so they cache that and retrieve the comments page count near real time.

    This isn’t a problem of numbers on the front page of reddit disagreeing with those on the comments page — which you might have noticed if you, y’know, looked at the screenshots.

    And if they did it in Lisp

    They did do it in Lisp. Then they rewrote it in Python. You might even have been able to figure that out from the text in this SOB entry if you bothered to read more closely.

    in PHP it would have failed b/c no great site works in PHP (wait…facebook, digg, etc).

    There are great sites written in PHP (such as Wikipedia). From the short list you provided, though, I’m not sure you know what a “great” site is. You must think “popular” and “great” are equivalents.

    Learn concepts about scaling web applications before spewing your spray about your favorite language.

    What’s my favorite language? About what favorite language of mine am I spewing? I’d really like to know, because I don’t recall “spewing” anything about “my favorite language” here. Perhaps you’re just a Python fan who leaped to the conclusion I’m saying dastardly, evil things about Python.

    I don’t personally like Python, but I know it’s a good language. It’s just not the best language for me. Try reading what I said without always looking for a fight with people, looking for excuses to believe they’re slandering your favorite language — and maybe you won’t catch yourself “spewing your spray about your favorite language.”

    Comment by apotheon — 5 June 2008 @ 09:25

  4. Well, I don’t know about post-upgrade, but pre-upgrade it definitely scored comments differently (for instance) based on the category (hot, new, controversy). I think it’s always seemed a little odd — I finally assumed their system was scoring things in its own manner. Since there’s no visibility into their sorting algorithm, it’s impossible to tell.

    Comment by Brandon — 5 June 2008 @ 09:47

  5. Hmm. I see that the numbers change if I change the sorting category.

    . . . though that doesn’t seem to solve the problem of numbers disagreeing on the same page. Both numbers — the number attached to the headline, and the numbers in the box to the right — change depending on which sorting category I view. In some categories, they match up, but in others they may not. In fact, looking at the sorting categories for the Copyright, Metallica, Microsoft, Napster, NIN, Prince, and Radiohead headline now, the “hot” numbers match up, as do all the others except the “old” category. In “old”, it shows a score of 4 on the left, but (3 – 2 = 1) in the box on the right:


    This problem existed before the most recent upgrade, too. In fact, a friend and I talking in IMs would often note that a single page would have different scores when viewed simultaneously by two different people online in locations thirty miles or so away from each other. This is part of what makes me think, as I indicate at the end of the SOB entry above, that it might be some kind of server replication issue. I guess I could have mentioned that earlier, but I didn’t think of it at the time.

    Comment by apotheon — 5 June 2008 @ 10:19

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

All original content Copyright Chad Perrin: Distributed under the terms of the Open Works License