Difference between revisions of "User talk:Tmtoulouse"

From RationalWiki
Jump to: navigation, search
(DynamicPageList: LiquidThreads!)
Line 160: Line 160:
::Can you request a backup at any time or is it just at scheduled times? Because if you can just click a button to copy the whole virtual machine updates are going to be a breeze. --&nbsp;[[User:Nx|<span style="color:teal">'''''Nx'''''</span>]]&nbsp;/&nbsp;[[User talk:Nx|''talk'']] 21:55, 3 November 2010 (UTC)
::Can you request a backup at any time or is it just at scheduled times? Because if you can just click a button to copy the whole virtual machine updates are going to be a breeze. --&nbsp;[[User:Nx|<span style="color:teal">'''''Nx'''''</span>]]&nbsp;/&nbsp;[[User talk:Nx|''talk'']] 21:55, 3 November 2010 (UTC)
:::Your allowed 3 backup slots, 2 for scheduled time one for whenever. So basically a daily, a weekly, and then a push a button and boom. tmtoulouse 22:22, 3 November 2010 (UTC)
:::Your allowed 3 backup slots, 2 for scheduled time one for whenever. So basically a daily, a weekly, and then a push a button and boom. tmtoulouse 22:22, 3 November 2010 (UTC)
::::Also I finally fixed the bug in LQT that was one of the two things making it unsuitable for general use, namely rollback and mark as patrolled didn't work (apparently that didn't stop them from deploying it on several WMF wikis...), so I think we can install that too, but not enable it by default, so if anyone wants to use it on their talk page, they can. --&nbsp;[[User:Nx|<span style="color:teal">'''''Nx'''''</span>]]&nbsp;/&nbsp;[[User talk:Nx|''talk'']] 00:22, 4 November 2010 (UTC)

Revision as of 00:22, 4 November 2010

Archives for this talk page: <1>, <2>, <3>, <4>, <5>, <6>, <7>, <8>, <9>, <10>, <11>, <12>, <13>, <14>, <15>, <16>, <17>, <18>, (new)

Am I going to get a response?

Will you permanantly ban me from this website? (talk) 15:31, 10 August 2010 (UTC)
No. tmtoulouse 15:42, 10 August 2010 (UTC)
And why so principled all of a sudden? You're prepared to stand idly by while Gerard brings about the wikipedian revolution here, why not solidify the revolutionary movement by removing your bogeyman? The thorn in your side? I want to leave this site, sooner rather than later, and would appreciate a rangle block and a 404 server ban. Will you please do this one favour for me. (talk) 15:59, 10 August 2010 (UTC)
tmtoulouse is a libertarian at heart. He would rather do nothing in the name of being principled than put any thought into being proactive.--Llort (talk) 16:27, 10 August 2010 (UTC)
Yeah I have never done anything proactive for this site or its users. tmtoulouse 02:38, 11 August 2010 (UTC)
I remember that time when Tmt pulled my fiancee and I from a burning car. AceX-102 02:40, 11 August 2010 (UTC)
And I remember the time he squashed a scary fucking spider for me. Trent is a brave man like that. Star of David.png Radioactive afikomen Please ignore all my awful pre-2014 comments. 02:47, 11 August 2010 (UTC)
And when he lanced the giant pustule on my neck. Not squeamish at all is our Trent. AceX-102 02:49, 11 August 2010 (UTC)
Don't forget the time he fended off a mugger for me! Trent learned karate for just such an occasion. Star of David.png Radioactive afikomen Please ignore all my awful pre-2014 comments. 02:54, 11 August 2010 (UTC)
You've got to admit, it was awfully decent of the mugger to wait for him to finish. --Kels (talk) 02:58, 11 August 2010 (UTC)
Better to be like Wikipedia than an Encyclopedia Dramatica wannabe. Plus I find it hilarious that I can provoke such a juvenile response out of you just by actually contributing. Oh well, you can't please everybody. And that's all the more I'm going to say on this. I feel as if I've feed the troll too much already Troll IAMELIPHAS (talk) 13:08, 11 August 2010 (UTC)

Template help

Hi! I'm pretty awful at wiki stuff. If I were to rename the template empiricism, say philosophy of science, would I need to go to every page using that template and change the name? sorry, pretty amatur question.--CentimeterINCHES 21:11, 13 August 2010 (UTC)

If you leave it as a redirect to a new template it will still transclude properly, however, it is best to change the references to the new template, particularly if it is not too many. You can find them using the "what links here" options, such as http://rationalwiki.org/wiki/Special:WhatLinksHere/Template:Empiricism. tmtoulouse 21:26, 13 August 2010 (UTC)
Thanks!--CentimeterINCHES 21:48, 13 August 2010 (UTC)
PS- Nobody will mind if I go ahead and change it, right?--CentimeterINCHES 21:58, 13 August 2010 (UTC)
For templates it is best to have a simple name that people can remember, especially one that doesn't have spaces. This isn't a software issue, but for convenience. If you plan on a rename/revamp, it would probably be best to try to make a new one, either in a sandbox of your own or on a subpage of the template first. Scarlet A.pngpathetic 10:45, 14 August 2010 (UTC)

Vadal Brake

I found what seems like a bug in the vandal brake, autoblocks never expire --Sir Onion Kneel before my vegetable might! 20:53, 23 August 2010 (UTC)

That's by design. -- Nx / talk 20:54, 23 August 2010 (UTC)
The method behind the madness is that if an IP is used by a vandal/spambot, then it will always be used by a vandal/spambot so expiry isn't necessary. Scarlet A.pngpathetic 20:56, 23 August 2010 (UTC)
s/method/shoddy, ridiculous justification that has no rational basis but helps us loft our noses a little higher in the air/ and you are much nearer the truth. Time drug Phantom Hoover! 20:59, 23 August 2010 (UTC)
(ec) No, the method behind the madness is that a binned user can request to be paroled - it won't happen automatically after a set time, only when they demonstrate that they are no longer a vandal (which sounds nice in theory, but epic fails in practice). Therefore the bin does not support expiry, therefore autobinned IPs are binned infinitely. -- Nx / talk 21:01, 23 August 2010 (UTC)

Meh, with out the historical context for the development of the tools it is sometimes easy to go off the deep end with conspiracy theories.

Really, the main focus of the vandal brake was the initial user edit braking. The idea being that new users would have their editing rate slowed to a level that shouldn't bother a normal editor, but would disrupt massive vandalism. Take a look at the number of pages this vandal hit on RW compared to CP.

Once a user reaches a certain age and number of edits this restriction goes away. If such a user were to go on a vandal spree then the brake wouldn't do anything, and hence why the vandal group was created.

When this tool was created our approach to vandalism and vandals was very different than it is now, and the tool as created was a better fit for that time.

Our approach has changed, and while the initial user brake is still handy the vandal group should probably be replaced by judicial use of short blocks. But because if historical contingency and institutional momentum the vandal brake is still often used.

To attribute it to malice or pretension is in one stroke both giving us too much and too little credit. tmtoulouse 21:06, 23 August 2010 (UTC)

Wasn't one of the reasons that blocked users could not protest their block because the ability to edit their talk page was either not available or you were not aware of it? -- Nx / talk 21:10, 23 August 2010 (UTC)
A problem which is easily solved by putting a warning in the block screen --Sir Onion Kneel before my vegetable might! 21:14, 23 August 2010 (UTC)
We were aware of it, but it was removed during one of our attempts to get rid of TK. Again, this was a time when there were maybe, 15-20 active users on the site, all old guard CP veterans and TK was about as power a lightening rod as you could get. We blocked him but allowing him to continue editing his talk page didn't change anything since it still continued the back and forth crap we were trying to purge.
This was also the first and only time I tossed in a 404 block on the server. That didn't go down well, and the compromise was to get rid of the block users editing their talk page and replace it with the vandal brake system.
This was also at a time when we often tried to purposefully engage our vandals in conversations, it was in general, a different zeitgeist. Things have changed, not really better or worse, just a different site. But institutional momentum is strong. I have being and advocate of the judicial use of 1-2 hour blocks as a means to stop vandalism for a while, but haven't gotten around to trying to curb the vandal brake as a tool. tmtoulouse 21:15, 23 August 2010 (UTC)
The problem is, whenever a "vandal" who finds an article that's less than favorable about them/their woo/their favorite writer etc. blanks an article or section, they are immediately binned, which doesn't exactly help with the whole engaging the vandal thing. -- Nx / talk 21:29, 23 August 2010 (UTC)
I agree that it is not something we really do anymore, personally I would do it very differently, for only a single incident of non-spam "vandalism" I would attempt to engage the user on their talk page. If that doesn't work or their are multiple attempts with no feedback then issue a 30-60min block and engage them again on their talk page. Most of the time that will either stop the vandalism or start the lines of communication. However, I am not here the majority of the time, and have personally dealt with very little in the way of vandalism in the last year or so. Maybe it is time to review our procedure for dealing with vandalism, with specific instructions and an eye towards moving things back towards dialogging rather than expelling. tmtoulouse 21:59, 23 August 2010 (UTC)
I'm not convinced the majority of wandalism is like that, however. If someone has an issue and wants to discuss it, they usually do. Page blankings and "HUURR!!! RW IS TEH GAYZZZ!!!112" wandalism is usually once-only drive by - we neither lose nor gain from the VB because they normally don't come back. I also haven't seen anyone drop a user or BoN in the bin for that sort of wandalism for a long time (I tend to wait to see if it's a clear repeat offender before using it, and issue a short-time block to calm it down while I find the bin link). Scarlet A.pngpathetic 22:44, 23 August 2010 (UTC)
With serial blankers I prefer to just block for an hour or so, and perhaps nine hours if they return after an hour to continue their fun. Doesn't seem worth vandal breaking BoNs unless theres wandalism that goes beyond a single day. I do think though that it should be kind of exceptional to prohibit people from being able to edit their talk page while blocked. Worth a forum post to get some opinions? Concernedresident omg!!! ponies!!! 22:49, 23 August 2010 (UTC)
Well, the key problem is that there will be a range of ideas of what is supposed to be done. Some will say "that's not what RW does" or "freedom of speech" or something like that, while others will be more "it shouldn't be tolerated" or "Trent isn't paying a fuck ton of money to host this crap" or whatever. All quite valid opinions on the subject, I think. However, while we have choice, people will disagree about what to do and when and in what situations - thus, the only way to really get consistency is to remove one of the options; either ban OR bin, with both not available. Scarlet A.pngpathetic 11:20, 24 August 2010 (UTC)

tomorrow I'll miss you

Close your eyes, and I'll kiss you, Tomorrow I'll miss you, Remember I'll always be true...
And then while I'm away, I'll write home every day
And I'll send all my loving to you

All my lovin' I will send to you-ou-ou! ħumanUser talk:Human 06:42, 26 August 2010 (UTC)

I think I screwed up

I just noticed a screwup in one of my DPL patches for the forum. Can I please have my server access back to fix it? (sudo passwd -u nx) -- Nx / talk 20:42, 28 August 2010 (UTC)

Done, let know if there are problems. tmtoulouse 20:47, 28 August 2010 (UTC)
Thanks, fixed it. You can relock my account. -- Nx / talk 20:59, 28 August 2010 (UTC)

Why are rangeblocks disabled?

I was going to give this guy a 30 minutes timeout, but for some reason I can't. Why? I know you don't like large blocks, but it was only going to be fore 30 minutes. Especially since the IPs are all Chinese proxies. --Sir Onion Kneel before my vegetable might! 23:11, 28 August 2010 (UTC)

It is not going to do any good anyway, it takes him more effort to switch proxies than it does for us to block and revert. No need to fundamentally change our policy for one troll. tmtoulouse 00:51, 29 August 2010 (UTC)
So you think . What is going on at RationalWiki? (talk) 00:57, 29 August 2010 (UTC)

what are you?!

Turning into ken? deleting everything lol

Bayesian priors

Hi Trent - I have a quick question about priors, since I recall that you know quite a lot about the topic. As I understanding, generating priors for a binary event (P or not-P) is pretty easy; if we have no information whatsoever we can even just guess .5. But what about priors for a hypothesis that explain a given set of data (let's call it "D")? As I understand it, the space of possible hypotheses (H1, H2, ... Hn) that can explain D is nearly infinite, so our prior for any one hypothesis would have to be very close to zero. When more experimental results come in we can increase our priors for some of the hypotheses, but it seems like there would still be a nearly infinite number of hypotheses remaining, preventing the priors from rising significantly. How, then, can we use Bayesian inference in this kind of case? (In philosophy of science, I believe this problem is called "underdetermination.") Thanks for your input, Tetronian you're clueless 15:10, 30 August 2010 (UTC)

Not sure how detailed you want to get, so I will start off with just a few general points and if you want we can go deeper from there. Off the top of my head here would be the major points I would raise:
  1. The issue is really small priors, the degree to which this is a problem is a function of the amount of evidence gathered, and the relative strength of the support that the evidence gives to various hypotheses. So the easy case is the strength of the evidence, if the probability of our data given one or two of the hypotheses is really really high relative to all the other hypotheses, they will quickly differentiate themselves. If the data doesn't quickly differentiate hypotheses we can still be saved by just collecting a lot of evidence, if the data can not differentiate the hypotheses given the time constraints of data collection then Bayesian theory still gives us the right answer, that all the hypotheses are equally likely.
  2. This is only a problem if you assume uniform priors. There is no reason to assume that we have to use uniform priors. Why cant we assign relatively high probabilities to those hypotheses that we think are most likely and then relegate the infinite number of highly unlikely probabilities to small p-values?
  3. And last, we can constrain our Bayesian stats to comparing only those models or hypotheses we care about by using something like a Bayes Factor to get relative probabilities. In this case we are not getting any sort of objective statistical data given all possibilities in the universe, but still have a perfectly acceptable tool for differentiating the handful of models we maybe interested in. tmtoulouse 18:03, 30 August 2010 (UTC)
  1. I see what you mean, but my concern is that the space of possible hypotheses is nearly infinite so priors (assuming a uniform prior) will be approaching zero. In this case, evidence can raise/lower the probability of certain hypotheses only slightly since there are so many other possible hypotheses. But based on what you said in #2 and #3, that problem is pretty much moot since there are other ways of reducing the size of hypothesis-space.
  2. Yes, using Occam's Razor would seem to solve the problem. I have heard of formalizations of Occam's Razor that use Kolmogorov complexity to quantify how complex a hypothesis is. However, I struggle with the idea of Occam's Razor even though it seems so intuitively obvious because it is, ultimately, a generalization and an a very broad induction.
  3. I've never heard of Bayes Factors before, I'll definitely have to do some reading on them. That would certainly solve the problem of absurdly low priors. Tetronian you're clueless 19:43, 30 August 2010 (UTC)
The issue seems to be revolving around the concept of "logical probabilities" or "objective probabilities." Is there such a thing? Can we really assign an objective probability to a hypothesis or model? If we abandon the idea that what we are looking for is the "true" probability of a hypothesis, and instead recast things in terms of subjective probabilities a lot of these issues fall away. Subjective Bayesian inference is pretty popular these days because of a myriad of problems with logical probabilities. I don't know how many people really defend objectivist Bayesian inference.
I find a mathematical analogy an easier way to understand this. Lets talk about curve fitting. For any set of data that we wish to fit a curve, there is an infinite number of possible curves. However, this is only a problem if we don't want to constrain our curve fitting algorithm in any way. However, if constrain our curve fitting algorithm to using only a bare minimum number of polynomials needed to fit the curve with some sort of desired accuracy there is often a "correct" fit. Why constrain the fit? We gain several key advantages: parsimony, tractability, pragmatism, and a solution.
So essentially we are subjectively valuing these things over the infinite solution space. If we have a list of principles that we subjectively value we can constrain our hypothesis space. Sounding a bit like Kuhn yet? Kuhn identified a variety of subjective almost visceral things that science values when comparing hypotheses. We can use these things to constrain the hypothesis space, and it doesn't matter as long as we are not trying to find an objective probability.
At the very root of this whole issue is the idea of whether or not we can determine truth, or whether the value of our ideas is rather a degree of wrongness. Encapsulated as "all ideas are wrong, some are just more wrong than others." This is what allows us to constrain that hypothesis space through various means such as Bayes factors, or non-uniform priors, etc.
Ultimately the question of whether or not we can objectively justify those principles we are valuing in our prior construction seems to come down to the struggle of the logical positivist, particularly with Wittgenstein's Tractatus. Or even Gödel's theorems. There just doesn't seem to be away to create the axioms or rules for evaluation of ideas within a system, using that system. So we must always appeal to something outside a system in order to evaluate claims within that system. This is going far afoot of where we started I think. The short answer is most people doing Bayesian inference are subjectivist, and as such have no problem constraining priors based on just about any reason at all. tmtoulouse 20:07, 30 August 2010 (UTC)
My belief in objective probability was shattered after reading two pages of E.T. Janyes, and I don't subscribe to it anymore. I suppose the problem I'm having is justifying the methods used to subjectively value things in solution space. (In my first comment above, I assumed uniform priors because I didn't find any of those methods viable. But now I do know better.) I do need to read Gödel and Wittgenstein; up until now I was intimidated by the level of math.
So it seems like the problem I was having is pretty much non-existent, since there are methods to decrease the size of solution space. This problem is just justifying them. But for that I will need to do some more reading. Thanks Trent! Tetronian you're clueless 20:17, 30 August 2010 (UTC)
No problem, I found Wittgenstein to be relatively accessible, Gödel was a slow slog. One interesting thing to look into, is the difference between informative versus non-informative priors. In my models where I introduce an agent to a new learning problem there is often no justification for priors at all. In this case you can use priors that are based around the types of functions your hypotheses have taken that essentially reduce the "power" of the priors to nothing, and so the likelihood functions dominate. Hence, "non-informative" prior. tmtoulouse 20:22, 30 August 2010 (UTC)

Site notice

The sitenotice only appears for logged-in users, not casual readers of the site. Is that intentional? - David Gerard (talk) 16:58, 1 September 2010 (UTC)

Somebody must've clicked "mark as read for anonymous users" on Special:Intercom. Probably Trent, IIRC only crats can do that. -- Nx / talk 17:16, 1 September 2010 (UTC)

Site not working right

The site is loading at speeds of 200-1000 bytes per second and is almost unusable! I can't even log in right now! Fixit! (talk) 18:25, 12 September 2010 (UTC)

I think it's over now. Blue Talk 18:32, 12 September 2010 (UTC)
A please somewhere in there wouldn't hurt! tmtoulouse 18:33, 12 September 2010 (UTC)
Mm, seems to have been cured. 18:34, 12 September 2010 (UTC) SusanG Toast
What went wrong? --Sir Onion Kneel before my vegetable might! 18:35, 12 September 2010 (UTC)
Its a secret. tmtoulouse 18:36, 12 September 2010 (UTC)
Trent spilt beer on the server. SJ Debaser 19:00, 12 September 2010 (UTC)


Are you gonna move those minutes to their correct location? ħumanUser talk:Human 18:32, 13 September 2010 (UTC)

There all ready. tmtoulouse 18:35, 13 September 2010 (UTC)

Happy! Happy!

Goat pilaf.jpg

Happy birthday! (well, it is by my clock, not my fault if you Canadians lag behind). Hope you have a great day. So, where's the cake? --PsyGremlinTala! 23:56, 18 September 2010 (UTC)

Have some lovely goat pilaf! It's a party now! I'm sure we have some cake around here somewhere... Here on RW time has four corners, all holydaze last 48 hours! ħumanUser talk:Human 02:26, 19 September 2010 (UTC)
Did someone say cake?
Happy birthday! -- Nx / talk 08:57, 19 September 2010 (UTC)
Happy birthday chumley,
Love, England. SJ Debaser 10:14, 19 September 2010 (UTC)
Happy birthday, Trent! I hope the research is going well. Tetronian you're clueless 20:01, 19 September 2010 (UTC)
Happy happy birthday, Trent! --Sir Onion Kneel before my vegetable might! 20:57, 19 September 2010 (UTC)

Thanks all, its appreciated, that brain cake looks pretty good! tmtoulouse 20:54, 19 September 2010 (UTC)

That cake needs a bontal frolotomy! ħumanUser talk:Human 01:11, 20 September 2010 (UTC)
Birthday.gifParty animal.gifParty.gif Punky Your mental puke relief 01:13, 20 September 2010 (UTC)


Did you get round to dumping the database backups somewhere public? I wanted to have a nosey. CrundyTalk nerdy to me 22:37, 10 October 2010 (UTC)

The database backups themselves are not public as they contain personal information such as e-mail addresses, real names, and passwords. The content of the site can be found at rationalwiki:content dumps. If there was something you were particularly looking for I might be able to help with that as well. tmtoulouse 23:12, 10 October 2010 (UTC)
No that's fine. I wanted to have a poke around with the content. Cheers. CrundyTalk nerdy to me 08:51, 11 October 2010 (UTC)

Turning a trick

The main page says we should trick or treat you. So we've got all this goddamn candy left over and I already need to lose pounds. You want to come over and pig out on Reese's Cups and Twix? You can sleep on the sofa if you promise not to wake up the dog when you barf. Doctor Dark (talk) 02:23, 31 October 2010 (UTC)

I am enroute! tmtoulouse 05:11, 31 October 2010 (UTC)

I want candy too. Or I will pelt your wiki with eggs. ΨΣΔξΣΓΩΙÐWeaselly.jpgMethinks it is a Weasel 21:40, 31 October 2010 (UTC)


So I did some digging around. There's a version of DPL in MediaWiki's svn which has the 1.16 bug fixed. However, the workaround depends on a change done to MediaWiki after 1.16, so it still doesn't work with 1.16. So, how would you feel about using more recent versions of mediawiki (specifically, the wmf-deployment branch, which is what wikipedia uses, so it should be relatively bug-free) and updating more often? I'd like to keep us more up-to-date instead of backporting and hacking in new features, but I didn't want to bother you with it and I feared it would result in significant downtime (like our MW1.14 update, though that was atypical). Normally though it should only result in a short period of read-only mode, the only problematic part with our current server is backing up the database before an update. How would that work on the new server? -- Nx / talk 21:37, 3 November 2010 (UTC)

More frequent upgrades is a better POA then what we have now. Some social prep will need to be made though since the newer versions are a big change to the GUI. But basically if we can get the rwbeta site working right then I just say go for it. The backup on the new server is a much better system since the machine itself is VPS, so backup is a copy of the actual VPS state at any given time. So you can get a quick backup with out ever needing to do the usual of shutting down mysql service etc. tmtoulouse 21:49, 3 November 2010 (UTC)
Can you request a backup at any time or is it just at scheduled times? Because if you can just click a button to copy the whole virtual machine updates are going to be a breeze. -- Nx / talk 21:55, 3 November 2010 (UTC)
Your allowed 3 backup slots, 2 for scheduled time one for whenever. So basically a daily, a weekly, and then a push a button and boom. tmtoulouse 22:22, 3 November 2010 (UTC)
Also I finally fixed the bug in LQT that was one of the two things making it unsuitable for general use, namely rollback and mark as patrolled didn't work (apparently that didn't stop them from deploying it on several WMF wikis...), so I think we can install that too, but not enable it by default, so if anyone wants to use it on their talk page, they can. -- Nx / talk 00:22, 4 November 2010 (UTC)