Spring 2010 Meta Thread
This post is a place to discuss meta-level issues regarding Less Wrong. Previous thread.
- 's comment on Existential Risk and Public Relations by (16 Aug 2010 10:01 UTC; 12 points)
- 's comment on Disconnect between Stated/Implemented Preferences by (17 Mar 2010 4:32 UTC; 6 points)
- 's comment on The Cameron Todd Willingham test by (5 May 2010 3:20 UTC; 5 points)
- 's comment on Loleliezers by (1 Apr 2010 6:20 UTC; 4 points)
- 's comment on Open Thread: July 2010, Part 2 by (30 Jul 2010 10:19 UTC; 0 points)
- 's comment on But Somebody Would Have Noticed by (9 May 2010 10:31 UTC; -2 points)
I’ve decided to write polling support for Less Wrong. In particular, I’m going to make it possible to make polls that ask for probabilities, and let people choose to vote either anonymously or on the record. Expect a beta of some kind by 22 Mar.
My polling code for Less Wrong is now (mostly) complete. Check it out with ‘git clone http://github.com/jimrandomh/lesswrong.git’. The syntax for creating a poll in a comment is:
The first is a conventional multiple-choice poll, with the options in curly braces. The second gives radio buttons along a scale, with labels (such as Agree/Disagree) on the left and right. After you’ve submitted a ballot, it will show you the vote breakdown. There are also probability and number polls, for which you can get the mean and median. There is also a ‘raw data’ link, which provides a CSV file with all of the votes broken down by question and user (either their username, if they unchecked the ‘vote anonymously’ box, or a number), so people can do fancy analysis on the results.
The main areas that still need work are the templates (the results look a little ugly), error handling, and bug-testing. If all goes well, it should be ready to deploy on Less Wrong in another week or two.
What is the status of this attempt to add support for polls to Less Wrong please?
Mostly finished, but lost motivation due to lack of feedback—I heard that a test was set up on an internal Benton house server, but never saw it or heard what the results were, and other things took precedence. I plan to come back to this, but need a better test environment or the same thing will happen again.
I was going through the top-rated comments list, and found the several-greats grandparent—did this ever make any more progress?
Unfortunately not. The code from where I left off is still up on github, if anyone wants to pick it up. (The URL seems to have changed—it’s now https://github.com/jimrandomh/lesswrong instead of https://github.com/jimrandomh/lesswrong.git).
:) Thanks! I am excited.
I have officially missed my self-imposed deadline, due to having less free time to work on it than expected last weekend, but I have made progress and will continue until finished. (Setting a deadline was just a willpower trick anyways, and it did work for that purpose.) My next suitable chunk of free time is this coming weekend, so I hope to finish then.
jim, are you working in a public repository? I can’t find you on GitHub
Woot, I’ve committed and submitted my first code change to the LW codebase, a proposed fix for Issue 200, learning some novice GitHub and git skills in the process.
Since what prompted me to take this step was to improve the Anti-Kibitzer script, I think my next learning opportunity is going to be a closer integration of the AK within the main LW code base, maybe controlled by a user preference.
Repo at http://github.com/Morendil/lesswrong
Integral anti-kibitz? That sounds great!
Turned out to be almost anti-climactically easy. I couldn’t sleep and returned to the computer to pass some time, started to look into how it would have to be done, and it turned out to be very nearly trivial so I went ahead.
Implementation details: I’ve added a check box to the Preferences page, checked by default, under a “Kibitz options” section, with a label saying “Show commenter names and scores by default”. I plan to add a link to either a new comment on the old Anti-Kibitz top level post, or first write a new top level post and link to that, to serve as end user documentation.
Any suggestions for improvement while I can still make changes at lowest cost?
I’ll wait until tomorrow to commit and submit it for approval by Tricycle, because I don’t trust myself to write bug-free code, even if it looks trivial, at 3am. Especially without having written any unit tests (I’ll have to look closer into what tests there already are before I start writing new ones).
Nothing’s committed yet except on my local development machine, but when I push stuff it’ll be to http://github.com/jimrandomh/lesswrong .
just checked out your code. First of all, you might want this:
http://github.com/MichaelBlume/lesswrong/commit/f792e4458b3b0707b67437fd167cdb21af58f3c2
(tl;dr false->False in poll.py line 52)
also, how do I in fact create a poll? I checked both the comment and article fields.
Cool, thanks. I’ve got a repo up at http://github.com/MichaelBlume/lesswrong—nothing original yet, but you can see tricycle, peerinfinity, and the continuing work on reddit in one tree.
Is it possible to make it so that all of the comments in a thread appear on the same page, instead of having deeply nested comments appear on a separate page with a “continue this thread »” link? The current setup can make it hard to follow discussions, and to ctrl+f find things in a thread.
Recent articles have generated up to three levels of this, and since every comment of a certain level creates its own ‘continue’ link, there can be many, many clicks required just to follow a single thread.
I wholeheartedly support this change.
YES! Please!! I HATE that “feature” of this site. It is incredibly annoying to not have the whole thing on one page.
Seconded so enthusiastically.
I think the minimum karma for posting (20) is too low. My fear is not an influx of shoddy posts. Rather, ~20 karma is too low because many first posting attempts don’t go well even if the user could eventually make lots of positive contributions. Often, the work isn’t up to our expectations and we’ll vote the post down to −2 to −5 (or sometimes worse). But since votes on posts are worth 10 karma the author’s karma total just gets decimated. Thirty karma isn’t a big deal to posters with totals in the the thousands, but a −3 to someone with the minimum requirements to post puts them lower than where they started. The numerical representation of the effort they’ve put in participating and trying to understand the content here is wiped out. Add to the fact that publishing here is really intimidating to begin with and the fact that we can be kind of mean: there is a decent chance that low karma posters whose first posts don’t go over well won’t come back (or won’t comeback for a long time), like in a video game when you die and haven’t saved. I fear this will happen if it hasn’t already (I won’t single them out but looking back there are one or two cases where this may have occurred).
With a higher karma level the hit from a bad post can be comfortably absorbed and people won’t feel so bad, it’s like having to restart the level instead of the game. Plus waiting probably increases the chances the post will go over well. We could leave this up to the discretion of individual posters but it seems likely to me that many people won’t properly anticipate the reaction they will have if their karma hits zero or below.
Also, if a power or award is easy to get it just isn’t as exciting when you get it.
Math plugin for the wiki doesn’t work for several months already, which causes the error
to appear instead of formulas. The formulas show on the old pages only because they are cached and were not edited recently.
I previously reported this problem in November 2009.
On user pages, could there be an option to sort the person’s posts/comments by karma score instead of by recency? When I click on a username to find out more about someone, seeing them at their best can be more informative than just seeing their most recent contributions.
I suggest that a new open thread be opened whenever the current one reaches 500 comments, since that’s the highest number that can be displayed at once, unless a new one would be opened shortly anyway. (Shortly meaning in less than somewhere between a week and three days, depending on how fast these open threads are filling. )
Poll: Vote by voting one option up and the karma balance down.
EDIT: Changing the poll options in light of khafra’s comment, “in favor” → “New thread after 500 comments” (4 votes at the time of the edit); “against” → “no new threads” (0 votes) and adding options for other thresholds.
New thread after 300 comments.
karma balance
New thread after 200 comments.
karma balance
New thread after 200 comments with an explicit “Please do not make useless comments here; let’s stay focussed on improving LessWrong.com″ in the post proper.
Hm, I was going to propose that the threads continue on the 1st and 15th of each month (the 14th of February, of course), and be titled “Open Thread: Ides of March edition”, “Open Thread: Calends of April edition”, etc.. Because the classical Roman calendar is awesome.
That doesn’t allow for continued scalability in the face of the growth of this site, which is not going to slow down unless the admins try to intentionally slow down the growth.
Or maybe there are cool names for the 1⁄4 and 3⁄4 points of each month; dunno.
The May open thread is at 400+ now. Are we are still in favor of this? New open threads weren’t much good last month since they weren’t promoted. I suspect that will have to change for the new threads to be worthwhile.
New thread after 500 comments.
karma balance
New thread after 400 comments.
karma balance
I would amend this suggestion to opening a new thread when the current one reaches a number significantly lower than 500. On the last open thread, 19 of 20 comments in the first 500 were replies to other comments.
That’s a good convention.
First person that interprets the poll results to be in favor of a new open thread should get some free karma.
New thread after a number of comments significantly greater than 500.
karma balance
No new threads.
karma balance
Apparently at least two people “cheated” in the poll?
Time for that polling feature to be implemented, I suspect.
Karma balance.
Minor bug report: the post 37 Ways Words Can Be Wrong currently consists of five ways words can be wrong, numbered 1-5, followed by 32 ways words can be wrong, numbered 1-32.
LW is, implicitly, a learning platform for philosophy. The post+discussion format is very good, but IMO it would be well complemented by something like book club format, which is a longer continuous experience. (Yes I know book clubs often fall apart, but a big reason for that is that the members do not really like to read, and are trying to precommit themselves. After some time nature takes it due. Should not be a problem here).
What do you fellas think?
A sound idea- do you have any particular books in mind?
I’d recomend Frege’s Foundations of Arithmetic, as it seems companionship could easily aid comprehension here.
I’m going through AIMA on my own now (1.5th time) - that would be a good one. Some basic books on Game theory would be good (I like Binmore). Maybe even some Dawkins stuff (extended phenotype looks interesting and not everyone has read it). Basically I am thinking about more ‘basic’ books, relative to this forum, something along the lines of “books of knowledge” from http://yudkowsky.net/obsolete/bookshelf.html , they should get a wide enough audience to make it interesting (perhaps not those specific books in an attempt to avoid groupthink ;). Not sure if Ferge falls into that category, I still feel I need more logic to tackle stuff like that.
Typographical error: you need to drop the comma at the end of the link for it to work.
You might be interested: http://lesswrong.com/lw/2br/less_wrong_book_club_and_study_group/
I might Indeed (note time-stamp!) =)
thanks, fixed
As billswift says, the monthly open thread easily gets full of too much stuff to wade through. The recent comments page can alleviate that, but it would be useful to get a similar view of the comments for a single posting, i.e. in unthreaded most-recent-first order.
There is a thread somewhere for suggestions about the software that runs LessWrong, but I don’t know where it is, which is also a problem.
See http://wiki.lesswrong.com/wiki/Special_threads
It’d be nice to have a link to that on every page; maybe as part of the “Promoted | New | Top | …” bar.
I would’ve said in threaded oldest-first order, with only the first post in each thread initially shown.
Issues, Bugs, and Requested Features—I think I found it through recent comments when I found it.
New open threads aren’t being promoted once the previous one hits 300 comments. The defunct April 2010 open thread should also be demoted.
I’ve noticed a couple of times recently interesting discussion attached to a strongly downvoted article or comment. I just thought of a problem description: The display system does not distinguish between interesting/uninteresting items (articles or comments) and interesting/uninteresting threads; an interesting thread is at the mercy of its ancestors.
This could be fixed by allowing items to gain rank in sorting/filtering based on their replies or, more elaborately, a high-scoring reply to a low-scoring item should be promoted above it displaywise. I have no wonderful ideas about how to visually format this.
There should be a Clay Prize for solving the problem of threaded messages on the Internets.
I think it would be possible to use slrn or trn as a model, now that JavaScript means it’s possible to update a page about as fast as telnet—but I haven’t seen anything like that.
I don’t know whether there’s a technical reason it isn’t feasible, or if usenet is too low status to replicate.
When I bring this notion up, the most common reaction is that no one wants to read unmoderated discussions.
What’s lesswrongian for “concrete-bound mentality”?
Trn is the best method I’ve seen for making long discussions usable on-line.
The .newsrc (a record of your preferences and what you’ve read) is the most crucial part.
Usenet was set up (at least by the time I was reading it) to be read on a telnet screen. Your computer probably has telnet, and you’ve probably never heard of it. Telnet screens were ASCII-only, and they could load pretty fast, even on a 56K dial-up connection.
The result was that it made sense to display one article (post) at a time, and that meant that a record could be kept of which posts you’d read, and that meant you’d never see the posts you’d already read unless you’d asked for them. [1]
This meant that you’d never have to hunt around for new posts in a big discussion.
LW could have something like that feature just by adding “recent comments” at the original post level, instead of just having it for all the original posts simultaneously.
However, trn had more than that. It had commands for navigating up and down the tree structure of a discussion, and the ‘t’ command for displaying the tree structure. That was only useful for medium-sized discussions—a browser-based system could do much better.
When you opened a newsgroup (discussion community), it displayed a list with the authors and subjects of each post you hadn’t seen before, and you could mark which ones you wanted to see.
I didn’t use killfiles (marking posts to read satisfied me), but they added a lot of flexibility. You could tell the .newsrc to never show you anything by a given poster ever again. You could “thunderplonk”—never see anything by that poster, replying to that poster, or mentioning that poster. You could never see posts with specific words in the subject lines, posts which had been sent to more than some number of newsgroups.
Killfiles could also be used to make sure you did get shown articles with specific features, but there was less drama about that, hence the name.
Slrn added scoring to trn.
[1]I don’t think there was any way to keep such a record efficiently for browsers until JavaScript came into common use.
The .newsrc also let you switch clients, while keeping the same records of what had been read.
Can you write up a précis on the nature of slrn and trn for the benefit of the post-September crowd? Edit: I think it might be difficult to appreciate the suggestion not knowing how it will work.
The main reading screen in trn had a two dimensional graph of the thread at the top of the screen (one node per reply) which you could navigate with the arrow keys—easily going to the parent post, the first reply post, or the next or previous reply to the same parent post. There were keys to move up & down a page or a line of the current post. Space moved you to the next page of a post, or if you were at the end, on to the next unread post/reply, in thread order. Multipage posts were marked as read only once you’d got to the bottom. There was another key to mark them unread, or to skip the current item without having to read all the way to the bottom. The top level post in a thread was technically similar to the replies, unlike the blogging distinction between articles and comments.
There was a previous screen where you could select which threads to view, they could be listed one per line (lots of other options too).
You could enter commands to select or skip posts matching a subject or author, and if desired, everything underneath them in the tree. You could save these commands to run every time.
The parts about being able to see the structure of the thread at a glance and being able to quickly glance over a list of threads to see what’s interesting sound a lot like why I like Bungie.org’s WebBBS forums so much. At B.org I can say “This branch looks dumb” and immediately skip to the next one; at LW I often have to scroll through posts one at a time to find the end of a branch, and I often just give up.
You can click the “[-]” link at the top of any comment to hide it and all its replies.
Thanks; I’ll probably start using that. That can still “[leave] an interesting thread at the mercy of its ancestors”, though, which I don’t think WebBBS or (apparently) trn does.
That does sound like a better user interface.
slrn and trn had some nice features, but I think overall the benefit from the Usenet model is that it decoupled the presentation and user interaction, which was done by decentralized clients, with the storage and propagation, done by decentralized servers. The fact that the clients could evolve independently really let them become very good at what they did, rather than the “almost good enough” of most web discussions.
I, also, miss the decoupling you mention. On the other hand, the user interface design has influence on the community by way of what-is-visible and the length of wieldy posts; e.g. mail/Usenet encourage long posts with quoting.
I suspect such systems were most effective in a techie society like the old USENET, but I could imagine it might be done still. There should still be a browser interface, however.
Thank you for asking. I’ve been complaining about how much better trn was for years (and, more recently, saying that JavaScript would make browser-based trn possible), and no one had ever asked for the details.
My brother rediscovered USENET not long ago, and had been much impressed, so...
How about a threaded/unthreaded option for viewing replies?
That would make it easier to keep up with big threads, instead of looking for the new comments in each sub-thread.
The “recent comments” sidebar is about an hour out of date right now, and has been out of date for at least all of today, for me.
I’ve noticed that the Overcoming Bias sidebar is often out of date—and not behind by a specific amount, but rather stuck at a specific date, displaying “Am I a Sim?” as the most recent post there.
It’s up-to-date for me. Clear your browser cache?
It doesn’t happen all the time—I don’t see any problem right now, for instance. So that seems unlikely.
It’s not just you.
And it’s still the case.
We hope we’ve just fixed it—sorry about the delay.
(this comment is actually a test that should invalidate the cache and force the sidebar to update itself)
Hallelujah, it appears to be working.
Eliezer told me that his preference for new intra-month open threads was 500 comments.
500 is a logical cut-off point because that’s as high as the software allows you to display at once without extra loading. Since discussions continue indefinitely on Less Wrong posts, it makes sense to start the new open thread somewhat before 500 comments, so there is some room for discussion to continue on the old open thread without going over 500 comments.
Why is it that the Recent Comments and Recent Posts sidebars are populated through AJAX instead of being generated on the server side, considering that they’re not dynamically updated or anything?
(I was thinking of submitting a patch that would have the server send them pre-populated, but recalling this quote, I figured I ought to first see if I can find out the perfectly good reason why it was done this way.)
Since the sidebars are the same on every page, the browser can cache them separately and download them only when they are updated. Before April 2009 they were static HTML.
Were there tests done to find out if this actually results in a speed increase? As they can’t be more than a few kilobytes altogether (probably well within the size of a typical TCP packet), I wouldn’t be surprised if doing four XMLHTTPRequests and manipulating the DOM actually slowed it down, even if the results were cached.
I’m using Chrome, and the WebKit web inspector is telling me that each page view is spending 200-300 milliseconds in XHR. Not sure if that means it’s not caching them correctly or if that really is how long the JavaScript execution takes.
You’re right about it not being faster. I had assumed it helped with server load, but the commit message mentions something about Google indexing.
It seems like a better way to do this would be to check the User-Agent and only send the sidebar if the agent isn’t Googlebot. Or, if we want this to apply to other search engines, then we could still do it with JavaScript, but just include the content encoded in a
Is it possible to change my username?
What should be our convention for making top level posts for posting links vs posting them in the open thread?
My suggestion is that top-level posts should be made when there is at least a paragraph of commentary on the link—if the poster does not have that much to say about it, link it in the open thread.
I assume you mean “should not be made unless”? Surely you don’t suggest that a poster make a top level post even if it seems inappropriate to them for other reasons?
Surely I do—you are correct.
Self censorship is high already. Top level link post even without a description is okay because a comment with a description will invariably be written and voted up to top.
In the end, let the voting system do its job. Every time you post, you risk your karma. Top level post is basically raising the stakes.
How about a top-level “put cool stuff in replies” post every so often?
I’m imagining it as a home for anything good (links, books, video....) which isn’t necessarily going to lead to significant discussion. As with open threads, if discussion happens anyway, it can be a subject for a top-level post.
I suggest a new subreddit with a ranked list of links and self-posts. A consequence of the current convention is that links are fine when they are above a certain level of goodness and that barrier is extremely high. If anyone wants to experiment, you could try other Ted talks, though it will be hard to control for the amount of text you write to go with your post.
I just upgraded my RSS reader and the links of the entries in the recent-comments RSS feed aren’t working; they’re relative URLs and the reader isn’t handling this. Furthermore, according to The Feed Validator using relative links isn’t valid there anyway. Please change the feed to use absolute URLs.
Also, the URL in the element is broken and trying to refer to reddit.
Recents are broken again—judging by these:
http://lesswrong.com/comments/?count=25&after=t1_1ub8
http://lesswrong.com/comments/?count=25&after=t1_1ub7
...looks like 1ub7 is responsible.
Edit: Aha! It’s on
http://lesswrong.com/lw/20z/announcing_the_less_wrong_subreddit/
in the thread following
http://lesswrong.com/lw/20z/announcing_the_less_wrong_subreddit/1uak
Edit 2: The most recent comment to create a crash is
http://lesswrong.com/lw/20z/announcing_the_less_wrong_subreddit/1ub4
Edit 3: 1ub7 no longer causes any crashes.
Google Chrome seems to be having all kinds of problems with this site. I’m getting pages loaded in all-bold, lots of links bleeding HTML, background flowing into improper elements, etc.
ETA: version 4.1.249.1042 (public release) on Windows 7.
I was getting bleeding html, etc on occasion (ubuntu chromnium). Refreshing fixed it.
I’ve been having the same problem. I’m on Chrome 5.0.360.4 on Mac.
I see problems with 5.0.342.7 (Official Build 42476) beta, on Debian Linux unstable.
I’ve had no problems with any version (up to current 5.0.360.0) on Mac.
Which OS are you on? My 10.5.8 machine stops at 5.0.342.7 beta.
10.6.2
I’m not seeing any problems with the 5.0.342.7 beta on Mac, but I generally use Stainless.
Time for June Open Thread Part 4. If you are reading this comment, you should make it! If no one makes it by the time Part 3 hits 550 comments I will make it.
Quick related pair of questions:
How long does a post have to be to need a cut?
How long does a post have to be under the cut to justify clicking through?
A good ‘rule of thumb’ would be that the article should be no taller than your browser window before the cut, including the site’s header. (Obviously, different people have different size browser windows, but then, they also have different size thumbs). Following that rule, I would think even a single paragraph after the cut should be justified.
ETA: This comes from the print idiom of “before the fold” and “after the fold”.
Transferring from “The Cameron Todd Willingham Test” a meta thread I began:
I don’t think it is a good idea to invoke any sort of controversy without some specific novel point to make. I would not object were it just a thought experiment in an open thread, but good cause is necessary for a top-level post. As I said to Jack, I would hold top-level posts to a higher standard than “don’t see the harm”. By the novel-insight standard, “The Cameron Todd Willingham test” fails on two grounds:
The general idea is an obvious generalization of something already explored in some detail previously: examining the evidence associated with a criminal case and drawing an independent conclusion about the guilt or innocence of the defendant.
The post neither promises nor delivers substantial analysis on any subject.
It is my contention that what Kevin wrote does not show enough effort to justify a top-level post.
It is derivative of the Knox posts. But it asks us to do new things with our skills. In the Knox test the right answer was just that AK and RS have a very low P(Guilty) and RG has a high P(Guilty). This post asks us to compare the relative innocence of two people for two different crimes, with different circumstances. These exercises are important because right now all we can do is basically say “low” or “high” of “fifty-fifty”. I mean, yeah, people gave numbers but I’m skeptical they mean that much. We need to get a better sense for how probabilities vary at the extremes and this will help. [Aside: Frankly I think we should be doing this kind of thing way more often. Someone tried to do this for the probability of a nuclear attack in this months open thread and no one would even try to put a number on their current beliefs (so there couldn’t even be an extended discussion of the issue).]
Kevin also asks the interesting question about a Bayesian justice system which I suspect will lead to a lot of discussion.
And I’m not saying it deserves to be promoted. It doesn’t. But that doesn’t mean it deserves negative karma. A +5 or +6 would be fine.
Is the post a reasonable open thread comment? Yes. But throwing out the bare questions, questions unaccompanied by any serious attempt to answer them in detail or elaborate on their relevance to our purposes, is not a substantial contribution to the community.
Edit: Why should I care about assigning numbers in cases like these? Why do I need a better sense of how probabilities will vary at the extremes? Why should I expect to gain such a sense from this task?
And what would a Bayesian justice system even look like? It’s not obvious, and the post provides no guidance.
The comments fleshed it out, and that’s what I wanted. I reserve the right to ask questions via top-level posts, if I care enough about the question to risk karma on the asking. If it’s a sufficiently good question, the question and answers together represent the “substantial contribution to the community”.
But the comments did, and that’s what I wanted. I reserve the right to ask questions via top-level posts, if I care enough about the question to risk karma on the asking. If it’s a sufficiently good question, the question and answers together represent the “substantial contribution to the community”.
Is there a site policy for cross-posting at other blogs? Should there be one? Maybe an expectation of cross-linking at least? …If only because it looks like plagiarism when the same article comes up on google at different sites (and possibly under different names).
The standard Internet convention roughly matches U.S. fair-use laws.
Honestly, though, ask permission to repost for each individual post. It’s not hard, and it works.
For sure permission is required to repost someone else’s work, I’m mostly asking about reposting your own work. We’re all well within our rights to do that, I’m wondering what the best practice is, though. Should we encourage those with separate blogs to crosspost? Discourage? Expect cross-links? etc.
Personal opinion: a reason I find crossposts annoying is that I always feel I have to click through to the crosspost and see what the commenters are posting on the crosspost as well as the original post. That’s a potential downside to think about when considering a crosspost. Obviously crossposts do have advantages too! I personally am not going to trash a crosspost just for being a crosspost.
I’ll also echo RobinZ’s suggestion of adding from/to notices whenever you do decide to crosspost something.
I like “Crossposted from” and “Crossposted to” notices, but I know of no convention.
How difficult would it be to add RSS feeds for comments on specific posts? They would have to be more than latest-10, of course.
Two things which would make if easier to track communication in LW:
I’d dearly like to have a most recent replies to my comments so that I don’t need to check by hand.
Less important, but pleasant—a way of finding out which comments have gotten karma changes.
The mail / envelope button below your Karma Score / user name on the upper right of the screen shows replies to your comments in reverse chronological order.
Thank you.
March 2010 open thread part 3 is ranked lower than March 2010 open thread part 2 on the home page. The current thread should probably be ranked higher on the home page.
Part 3 isn’t on the front page—it hasn’t been promoted. The open circle with the karma number gets filled in when it is.
Which makes me notice: this hasn’t been promoted, either. That’s odd.
I’m following the lesswrong RSS feed via Google Reader. I just noticed that some top-level posts are not in my GR list. (Coincidentally, this one is an example.)
Did anyone notice something like this? Any ideas how to debug the problem?
The feed settings in GR show http://lesswrong.com/.rss as the URL it retrieves from.
That’s only for the front page posts. Use http://lesswrong.com/new/.rss for all new top level posts.
Thanks, that seems to work.
This is a comment.
This is a reply.
This is a self-referential reply.
This is a self-referential reply
You should change that link to look like this:
You mean, “This is a self-referential reply produced without going back and editing”?
This is not a pipe.
This is not a link.
Are you sure about that?
The empty string in the hostname position denotes localhost. So you have the same as http://localhost. The linkness of your link is syntactically present and referent-ly observer-dependent.
This is an invocation of authority. The URL in the grandparent post does not match the grammar specified in RFC 1738, and none of the browsers I have on hand to test with will follow it. Which browser sent you to localhost?
1738 is insufficiently formal, and AFAIK is superseded by 2616 and 3986. 3986 (URI Generic Syntax) says that the host component is required, but one of the alternatives for it is “reg-name” which may be zero-length, but:
RFC 2616 (HTTP/1.1) specifies the host as being non-optional and using the grammar from RFC 2396, which defines it such that the empty string is not included. Therefore my previous claim was incorrect per specifications.
Safari 4.0.4 Mac implements http:// and http:///foo as equivalent to http://localhost/ and http://localhost/foo/, possibly by analogy to file:///foo/.
FWIW, chrome opens an “about:blank” tab, not localhost.
This is twice the not a pipe.
[http://x This is not a link].
[http://]
[http:// This is not a link].
Cool. Surprised no one has painted an infinitely recursive version yet.
Tried, but got a stackoverflow, and stackoverflow didn’t help to solve it.
This comment is a statement of agreement with its first reply.
This comment is a statement of disagreement with its parent.
This is an off-topic reply.
Voted down for being off-topic.
Nonononono, you say “this is an explanation for a downvote of an off-topic reply”.
This, by the way, is a clarification of thread protocol.
This is meta-humor.