User talk:Gmaxwell

I prefer to keep discussions unfragmented. If I post on your talk page, I will notice any replies posted there. Unless you request otherwise, I will reply here to comments made here. I will usually post a brief note on your talk page to let you know that I have replied, unless your talk page instructs me otherwise. If you write a reply to me here, I may decide to move your text back to your talk page in an effort to keep the thread in one place. If you disagree with that, then please do not write to me here. If you are just pointing out something written to me elsewhere, edit here. Such pointers are useful if you've written to a comment I made many days ago.

Previous discussions:

• Archive 1
• Archive 2
• Archive 3

Roomba

Hi! A little bird tells me that you've made a server-side vandalism-detection tool with IRC output recently. I'm working on a client-side javascript tool and was curious about yours. Do you plan on adding some way of accessing it from a page on en.wikipedia.org, or any web interface? If so, given the power of the server side, I can probably stop working on my tool.... anyway, I'd be interested to see what yours does. Is there any code around that I can look at? Lupin|talk|popups 15:11, 12 November 2005 (UTC)

• Like yours it catches words likely to signal vandalism. It only catches them if the new version of the article has the word but the old versions didn't. It's in the vandalism channel on IRC. Phase I is what were working on now. Phase II with be adding in my bayes engine. Phase III will be working into an application that spreads the review load around between reviewers. I don't have a plan to make it web accessable by itself because I think it's of diminished value until I can spread the load around. (The thundering herd of people all trying to fix one article doesn't do much). You're welcome to the code but I'm avoiding wide public distribution, so I'll email you.

• At the moment, roomba catches additions of speedy notices to articles but it ignores the removal of CSDs. It would be good if it tracks them too. Tintin 15:49, 12 November 2005 (UTC)

• • Oh great idea, easy to impliment. We'll have that tonight! --Gmaxwell 22:16, 12 November 2005 (UTC)

• Thanks for the code and the list - I've added your words to my list. Lupin|talk|popups 23:03, 12 November 2005 (UTC)
  • BTW, I've moved my bot to #wikipedia-en-suspectedits Having it apart will make it easier for me to work on, and I suspect I have some phillosophical differences with the people involved with WP:CVU. ... Also, they won't remove the blanket tor ban. --Gmaxwell 23:44, 12 November 2005 (UTC)
    • Well I'd like to work on any philisophical differences there maybe, hopefully we are all working to the same end, other than the tor issue (which I'll try and find out more about, I know the reasoning but not how broadly supported it is) what are the key issues to you? As you say it might be easier for you to work on it separately for a while anyway but I'd hope we could bring the various efforts together at some point. Thanks --pgk^(talk) 11:04, 13 November 2005 (UTC)

• I've been thinking about the best way to add a web-accessible frontend to this tool. Can you output to a text file hosted on en.wikipedia.org? If so, you could send output to a file and I could write a script that polled it every so often and updated a list in more-or-less real time. (If you can only output to a file hosted on a different server then this can still be done, but not quite as neatly I think). Lupin|talk|popups 23:41, 18 November 2005 (UTC)

Regarding your fair use orphands list on the toolserver

You know http://tools.wikimedia.de/~gmaxwell/cgi-bin/report_orphan_fair_use.py

Very cool tool by the way, much better than sifting though the category or working of an offline list that get outdated after a couple of days... Anyway I noticed something odd. The links don't seem to match up with the titles, for example Image:3eb_blue.jpg actualy links to Image:Dutchcrown.PNG and so on, it confused me quite a bit untill I realised what was going on. Not a huge crisis since everyting have to be dealt with anyway, but you might want to fix whatever the problem is so clicking on a link actualy opens the "right" page. --Sherool (talk) 22:32, 14 November 2005 (UTC)

WOOPS! I'd just made a change before toolserver went down the other day so that every user gets the list in a different order, so that everyone isn't trying to delete the same content at once. But I forgot to fix the link as well. I've fixed it now. Thanks for pointing out my mistake! --Gmaxwell 23:13, 14 November 2005 (UTC)

The images in Category:Screenshots of Wikipedia, for example Image:010405ac.png, are not fair use, even though it's a subcat of Category:Fair use screenshots. User:Roomba needs to know about that exception. dbenbenn | talk 05:22, 15 November 2005 (UTC)

Thats not correct for vast majority of them, as they contain copyrighted browser UI elements. --Gmaxwell 16:27, 15 November 2005 (UTC)

If you'd like I can exclude these from the orphan report list so people will not delete them until we settle the issue. --Gmaxwell 16:30, 15 November 2005 (UTC)

IRC Bot

I really like your IRC bot, can you please host it in #wikipedia-en-vandalism channel? Or at least a dupe the data feed to there? I pmed you on IRC but there seems to be some sort of a problem. :) take care --Cool Cat ^Talk 17:39, 15 November 2005 (UTC)

I replied to you right when you asked. I told you that I'm still actively working on it, and I'm on the road for the next few days.. I'll be glad to talk about sharing a channel for it, although I'd like to see some changes to the operation of the other bot as well. --Gmaxwell 18:46, 15 November 2005 (UTC)

fair use bot

Template:Wikipedia-screenshot should not be tagged as unused fair use. Its GFDL/GPL. Modified ones are therefore under this license too. Justinc 12:26, 16 November 2005 (UTC)

Please read my reply to this above. This is now the third time I've discussed this on my talk page, fair enough I've since archived the first, but the second is just above.. --Gmaxwell 23:03, 16 November 2005 (UTC)

Sorry I didnt see that. Many of them dont include copyrighted elements. Please exclude them for the moment. Quite happy to police them by hand but it is not fair for people to be encouraged to delete free content. The category should only include free stuff, but that requires a lot of knowledge of copyright status. Have notified sone users that they should remove the copyright bits. The more judgements have to be made the more problematic and slower deletion is. Justinc 01:34, 17 November 2005 (UTC)

No problem on the missed thread, it's just that I've discussed it a few times with other people and this is the first time we've gotten beyond me pointing out UI elements. :) Sounds like a fine solution for now. I've adjusted the report so the category Screenshots of wikipedia are excluded. --Gmaxwell 02:23, 17 November 2005 (UTC)

Oh crud, it's not actually trivial to remove it from that report.. it's out of the orphan fair use report but not out of the tagged report page. I'm not sure if I have time left tonight to change that. --Gmaxwell 02:27, 17 November 2005 (UTC)

Got it. Time for bed now. --Gmaxwell 02:34, 17 November 2005 (UTC)

Video editing

Hey, I was wondering...can you help guide me in converting some files from the ASF and WMV file formats into the Ogg Theora format? I'm using Arch Linux, and I have mplayer installed. Thanks. Johnleemk | Talk 12:45, 16 November 2005 (UTC)

Ahh, if you're on linux it's pretty straight forward. Does your distro have a package for ffmpeg2theora available?--Gmaxwell 23:09, 16 November 2005 (UTC)

Unfortunately, it does not. I presume I can simply download the binary from the official website and install it on some corner of my hard drive? Johnleemk | Talk 07:11, 17 November 2005 (UTC)

Yes sir. [1] It's a quite simple tool. Let me know if you have any problems getting it going. --Gmaxwell 11:40, 17 November 2005 (UTC)

Well, it turns out that after converting about four seconds of any WMV file, ffmpeg2theora starts throwing the error "[wmv2 @ 0x859efe0]J-type picture is not supported". Johnleemk | Talk 14:21, 18 November 2005 (UTC)

Can you email me the file? I'll convert it for you and tell you what I did. --Gmaxwell 15:18, 18 November 2005 (UTC)

It was something of a pain to convert, but the result has been emailed to you. The file is a WMV8 file, and there is no open source decoder for this format. Mplayer includes an encoding module called mencoder which is able to use the Windows codecs directly as long as it is used on i386. I was able to use this to produce an uncompressed AVI, when ffmpeg to theora was then able to handle. When you have the chance please encourage people to use codecs which are less than completely proprietary! --Gmaxwell 21:08, 18 November 2005 (UTC)

Source

Hi, sorry for my poor english. What is the script that your bot runs? Can you send me a copy of it? --555pt 01:29, 17 November 2005 (UTC)

I'd be glad to send you any source you want, althought I am not the best coder. What part of the bot do you want? --Gmaxwell 11:41, 17 November 2005 (UTC)

I need specialy the part that add a template to unused images, to help my work in pt.wiki. And, if you have any code to detect and mark images that don't have description page, send me it too, please. --555pt 19:26, 21 November 2005 (UTC)

Bot Madness!

Hey, I'm working on a similar bot that takes a different approach (at least, from what I'm told), might be better to merge this techmique into your bot somehow... mine checks all links in the article against the SuRBL (a DNSBL of URI's commonly reported as spamvertised sites), and may, in the future check anon IP's against other DNSBL's as well.... anyhow, the technique is easy, just append the URL with .multi.surbl.org iirc, and resolve it, if it returns 127.0.0.x(1-16, iirc), then, it's been reported... (i.e. vilerage.us.multi.surbl.org , or whatever)... anyhow, I've regged #wikipedia-en-spam, if you'd like to sit your bot there, seems more appropriate than the vandalism channel ;] I know i'm all over the place tonight (too much caffiene) but, weighting a rBL responce might be helpful in determining the spammieness of a host, or url ;] Anyhow, good work, have fun ;] If you would, please at least notify me of replies on my talk ;] --VileRage (Talk|Cont) 10:19, 19 November 2005 (UTC)

I can put that into my bot I already extract external links, just don't do anything with them yet. I'll probably add this today. --Gmaxwell 17:10, 19 November 2005 (UTC)

Cool, dunno what lang. you wrote your bot in, but mine's in PHP, if you want the code i use for that process, lemme know ;] --VileRage (Talk|Cont) 06:29, 20 November 2005 (UTC)

Roomba

Bot down! Bot down! — BRIAN0918 • 2005-11-21 19:49

Award

For flooding IRC like no other, I award you this surreal barnstar. Redwolf24 (talk) Attention Washingtonians! 03:08, 22 November 2005 (UTC)

Hey

I want to apologize if I came off as an idiot in the Marbled Orb Weaver nomination page. Sometimes, I begin to lose my tact, and don't know when to shut up! The only thing I was trying to say was that, assuming both images are viewed at actual size, the downsampled image will appear smoother and sharper. In my experience, this has always proven true, and this has even been mentioned at places like dpreview.com. If I'm wrong, well, I suppose I can do the man thing and apologize. I wasn't angry or anything, and I want to apologize if I came off as being overly arrogant(possible) or dumb. (Very possible).

Sigh. It seems this happens everywhere I go. I hope I haven't damaged my [reletively] good reputation on Wikipedia too much. Well, there's always the two week rule.PiccoloNamek 01:29, 24 November 2005 (UTC)

No need to apologize. I'm the one thats been harsh. I think you've been ignorant, and perhaps you continue to be ignorant... but ignorance is curable. :) I know what you're saying, but you're making a completely useless comparison. When comparing images we must compare them at the same *size*. If I made two images, one slightly out of focus and at ISO1600, and another perfectly focused and at ISO100... but you viewed the first while standing across the room and the second while right at the screen, you would probably decide the first was sharper and less noisy than the second. That doesn't make it better. Mediawiki decouples resolution and size. We can have any size we want. It's best to upload the highest resolution. --Gmaxwell 02:13, 24 November 2005 (UTC)

Good God! So that's it. I think this is for most forehead-slappingest moment in my entire life. Ok, let me get this straight: What you're saying is that the 2560 image and the 1280 image will look the same when thumbnailed at say, 250px, for example? Right? (I hope I'm right.) It isn't that I'm wrong, it's that I'm arguing from a completely different perspective than you. I'm arguing from the perspective of someone who rarely spends time looking thumbnails. If I see an image and think for even a second it might be interesting to look at, I'll open a new browser tab and go into the actual size view, and, unless there is something severely wrong with either my eyes or occipital lobe, the downsampled image DOES look better when compared to its unedited brother. I can verify it with almost every image I've ever taken. Reducing the image size by 50% means you're using 4 pixels to calculate 1 pixel in the new image, and this drastically reduces the noise in the photo, reduces the appearance of JPEG artifacts (when I actually use JPG), and gives the image the appearance of being sharper. But I fully understand what you're saying now, that doing that isn't necessary for the purposes of Wikipedia, because it will look the same in a small thumbnail as the downsampled image. The problem was, I wasn't arguing from that point of view, nor did I consider that in the optimization of my images. (I work on them for optimal apperance at actual size, and in most cases, actual size is usually 1280x960). I THINK I understand what you're saying now, it's just that the language just wasn't being processed properly. (Thufir Hawatt: "Clearly, word-sounds were not being linked up here in the normal manner.".) I HOPE that was right, or at least partly right.

Better, but I'm not just talking about thumbnails. I'm talking about any case where the two images are the same physical display size. I.e. if they are both 250px wide thumbs they are the same physical display size (presuming both are viewed on the same screen). I'm just as equally talking about the case where you display the images so they fill your screen, or if you print them both on the same size piece of paper. These are the only situations you can use to rationally compare two images. In any of these cases your downsampled version will either look the same, or worse than the higher resolution image. Which one of these possibilities will depend on the resolution of the display medium. I think it's funny that you reply that you don't spend much time looking at thumbnails when your initial response to my complaint about the resolution of your image was that "being print-worthy isn't a criterion for becoming a featured picture", implying we should be happy with on screen thumbnails. Featured pictures are about being the best pictures we have to offer, I think being print worthy is a critical factor, others might not agree but very few would say it was unimportant.

Your claim that downsampling reduces noise is both correct and completely false. If the ultimate display medium's resolution*size is smaller than the downsampled image both will look *the same* (modula quality lost due to the extra processing on the downsampled image, and no .. bicubic isn't the end all of resampling, see this). If the display medium's resolution*size is larger than the downsampled image then the downsampled image will be less noisy, but it will also be blurry. You could achieve exactly the same thing with a gaussian blur and leaving the picture size the same. Would you really advocate everyone run a 2px gaussian blur on every image before they upload? Try it, downsample an image 50% then back to the orignal size, and compare it to a 2px gaussian blur, they should look pretty simmlar, since there is about the same amount of information loss and it's basically the same process (it uses 9 pixels to compute one). Oh and by the way, your bicubic 50% downsize uses 16 pixels of the source image, not 4. Bilinear interpolation would use 4, and that often looks like crud. --Gmaxwell 04:50, 24 November 2005 (UTC)

As for the joke in the unripe lemon nomination, the fact that I bolded and italicized the word "extremely" should have clued most people into the fact that it wasn't an extremely advanced technique at all, but rather, and extremely easy one, in that case, a basic contrast mask.PiccoloNamek 04:11, 24 November 2005 (UTC)

No, it wasn't clear to me or to other people. You keep coming off with such arrogance that it's hard to see more of it as a joke. For example, "I know my way around Photoshop a hell of a lot better than most people." or the other comment I pointed out in my response on how you have trained yourself to detect image problems. --Gmaxwell 04:50, 24 November 2005 (UTC)