Jump to content

Wikipedia:Village pump (technical)

From Wikipedia, the free encyclopedia

This is an old revision of this page, as edited by Finlay McWalter (talk | contribs) at 22:30, 7 September 2005 (→Newuserlog: contribs link). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.

 Policy Technical Proposals Idea lab WMF Miscellaneous 
The technical section of the village pump is used to discuss technical issues. Bugs and feature requests should be made at BugZilla since there is no guarantee developers will read this page.

FAQ: Intermittent database lags can make new articles take some minutes to appear, and cause the watchlist, contributions, and page history/old views sometimes not show the very latest changes. This is an ongoing issue we are working on.

Details about the occasional slow speeds and deadlock errors: here

Please sign and date your post (by typing ~~~~ or clicking the signature icon in the edit toolbar).

Please add new topics at the bottom of the page.


Technical archive

Discussions older than 7 days (date of last made comment) are moved here. These discussions will be kept archived for 7 more days. During this period the discussion can be moved to a relevant talk page if appropriate. After 7 days the discussion will be permanently removed.

Make history less cluttered by reverts

Currently a large part of the history is cluttered by vandalism and reverts thereof. This is inconvenient when I want to simply look at the history in order to (for example) see where some inaccurate fact comes from. I think there should be a way for the history UI to show identical versions in the history, and a way to hide all the versions in between (in edit wars this would only help those agreeing with the current HEAD version, but it does no harm anyway). Of course, something similar to "svn blame" would be even better. R6144 05:14, 25 August 2005 (UTC)[reply]

An excellent idea. It would also be lovely to be able to see when a particular bit of text worked its way into an article (i.e. without needing to perform the manual binary search that I currently use). Dmharvey File:User dmharvey sig.png Talk 17:25, 25 August 2005 (UTC)[reply]

What is "svn blame" ? StuRat 21:02, 26 August 2005 (UTC)[reply]

At a guess its a system for subversion for figuring out what edit is responsible for introducing a particulat peice of text. Plugwash 19:54, 31 August 2005 (UTC)[reply]
"svn blame" is similar to "cvs annotate". For each line in the file, it tells you the time and/or version number, as well as the author, of the most recent change that affected that line. —AlanBarrett 09:30, 5 September 2005 (UTC)[reply]

Is my password vulnerable?

Yesterday, someone I know told me that he uses Wikipedia but won't tell me his username because there's a vulnerability I could use to get his password, and that I could find more information in Bugzilla. He's very knowledgeable and qualified to make this statement, but also quite paranoid, so I don't know how seriously to take his statement. A Bugzilla search for "password" turns up nothing like this. Any comments? ~~ N (t/c) 04:36, 28 August 2005 (UTC)[reply]

Just the usual PW advice:
1) Don't use anything obvious, like your username or real name.
2) Don't use English words.
3) Use a combo of letters and numbers.
4) Make it as long as possible.
5) The "remember password" option does store it as a cookie, which makes it more vulnerable.
StuRat 04:57, 28 August 2005 (UTC)[reply]
Yeah, I know, this question is about a technical bug in MediaWiki. And I'm pretty sure the "remember password" option only stores a session ID in a cookie, which would allow someone else to use my account (temporarily) but not read my password. ~~ N (t/c) 05:01, 28 August 2005 (UTC)[reply]
We used to send out the MD5 hash of the password in a cookie, if you selected "remember me" checkbox as you logged in. This, combined with a few XSS vulnerabilities that we hadn't bothered fixing, is probably what led to the compromise of some passwords by a brute-force attack on the hash. All the XSS vulnerabilities are fixed now (we hope), and contrary to what StuRat says, we no longer send out the password hash under any circumstances, so arbitrary HTML is not really a problem anymore. We've also had some very weak passwords (most often blank) broken by dictionary attacks against the server. This attack is only feasible if the target password is one of the few hundred most common passwords used. -- Tim Starling 05:17, August 28, 2005 (UTC)
Well, how about an SSL login then? That's still a major sniffing vulnerability. -- Curps 06:26, 28 August 2005 (UTC)[reply]
Not one that's specific to MediaWiki. Our cluster design makes SSL difficult to implement, although we may do so concurrently with m:SUL support. JavaScript password hashing would be much simpler and would provide reasonable security. -- Tim Starling 08:21, August 28, 2005 (UTC)
I hope it can be done one way or another, reasonably soon. Sooner or later, vandals exploit every vulnerability they can find. -- Curps 08:34, 28 August 2005 (UTC)[reply]
Vandals are teengers who barely know how to use their own browsers, maybe you mean hackers. And it's only an issue if your network is insecure on your end, which is increasingly rare these days. It's far more likely that a breach will come via unauthorised code running on the user's computer -- we often get spam posted from legitimate user accounts due to spyware infections. -- Tim Starling 10:26, August 28, 2005 (UTC)
It's perhaps a little too complacent to assume that all vandals are clueless. Wikipedia is a very high-profile target and seriously compromising it would be worth major bragging rights. Passwords could be sniffed at any hop, no? Maybe some vandals have day jobs. It would be nice to have SSL. -- Curps 11:57, 28 August 2005 (UTC)[reply]
Not to mention, a lot of people probably use the same password for WP as for everything else, so cracking one could be quite profitable. ~~ N (t/c) 17:01, 28 August 2005 (UTC)[reply]

Worst possible scenario is that a Sysop's account get's compromised and then the vandal uses a bot to wreak all sorts of administrative havoc on the project. Of course, this is a problem for any system... — Ambush Commander(Talk) 17:12, August 28, 2005 (UTC)

Nickptar, please ask your friend to e-mail me with details of this vulnerability. If there really is one, obviously we'd prefer to fix it. (He can PGP-encrypt the mail if he likes; I PGP-sign the release announcement mails so he can do a sig check against one of those to confirm that I really am the MediaWiki release engineer.) --Brion 03:54, August 29, 2005 (UTC)
Update: all he's heard is that there are widespread rumors of such a vulnerability, which he found out about from a Slashdot comment. Doesn't sound very credible. ~~ N (t/c) 22:17, 31 August 2005 (UTC)[reply]
More info: the /. article was about a month ago and linked to a Wikipedia article. Some clueless person saw that the linked article had been vandalized and thought WP had been hacked. This started a serious discussion of WP vulnerabilities and prompted the comment about the password issue. Hopefully a regular /. reader will remember this. ~~ N (t/c) 01:44, 1 September 2005 (UTC)[reply]
  • Any website that requires password login and doesn't use SSL is pissing in the wind. If you have to type in your password during a normal login, anybody who controls a computer between yours and the host computer can see your password as plain as day. It passes unencrypted and they just write it down alongside your username. For practical purposes, on sites such as Wikipedia, this is not a problem. But if you ever pass sensitive information across a network that you do not totally control you may want to rethink what you're doing. Even SSL isn't uncrackable, but it's better than walking around with your pants around your ankles, which is how most web transactions are carried out. --Tony SidawayTalk 02:07, 1 September 2005 (UTC)[reply]
    • IMO this is overly alarmist. If SSL (really TLS) is properly implemented, it's very close to uncrackable. Various browsers' implementations of SSL have been cracked (seems like this might be worth an article, off hand I can't find one), but I think all the current versions (when using 128-bit or greater keys) are essentially uncrackable. Of course, most users don't understand enough to know whether or not to trust a "secure" web server (many such servers shouldn't be trusted), but I think claiming SSL is crackable in general simply isn't true. -- Rick Block (talk) 04:13, September 1, 2005 (UTC)

How can WP use Squid?

I know that Wikipedia uses the Squid cache to cache some outgoing webpages, but I don't see how this can work. I mean, when I view an article, I see all kinds of things that I get to configure: the skin, my username and links at the top, configurable dates, thumbnail sizes, etc. How can the page content be cached but these not? ~~ N (t/c) 18:44, 29 August 2005 (UTC)[reply]

Stuff for signed in users isn't cached by the squids. Only anons get cached pages; as the overwhelming number of pageviews are from people who aren't signed in, this turns out to be a very effective optimisation. -- Finlay McWalter | Talk 18:50, August 29, 2005 (UTC)
however the page content is cached in the parser cache even for logged in users. Squid is just used because it can serve up the anon pageviews with less overhead than the parser cache system. Plugwash 18:56, 29 August 2005 (UTC)[reply]
Cool idea, thanks. Is it the wikitext or HTML that's cached in the parser cache? ~~ N (t/c) 18:59, 29 August 2005 (UTC)[reply]
The HTML from article content is cached. — Sverdrup 19:13, 29 August 2005 (UTC)[reply]
Then how are date formats, image thumbnail sizes, and the like configurable? From my brief experiments, those are done through changing the article HTML. ~~ N (t/c) 22:37, 29 August 2005 (UTC)[reply]
I think what it actually caches is the tree of objects obtained by parsing the wikitext but i'm not totally sure. Plugwash 23:03, 29 August 2005 (UTC)[reply]
Those options that change rendered HTML are included in the hash key that cached output is stored under, so users with the same options share cached data with one another. --Brion 23:33, August 29, 2005 (UTC)
Ah-hah! Thank you. ~~ N (t/c) 23:38, 29 August 2005 (UTC)[reply]
Does the parser cache explain the "memcached" apaches on the server list? It appears the large-memory apaches are configured differently. – Fudoreaper 00:52:19, 2005-09-01 (UTC)
No, memcached's cache database hits. --fvw* 01:13, September 1, 2005 (UTC)
I beg to differ. Most of the memcached space is used by the parser cache. It also stores various other objects. Using it to store database results in the naive manner suggested by the memcached webpage would be a bit of a disaster for MediaWiki. -- Tim Starling 02:11, September 2, 2005 (UTC)

Bots

Where should I go to ask questions about bots. I can't find software to build a bot, since pywikipedia only currently supports a handful of wikis. Did I miss something or will it work with any MediaWiki wiki? Also, would I just be able to program a bot into a php script (Will I be able to run a php script to perform an automated task?)

Thank You, shardsofmetal 21:35, August 29, 2005 (UTC)

Regarding which bots do what, you should probably ask on the wikitech-l mailing list. If it's your wiki, and you have console (and database) access then you don't need a bot at all - you can just write a script that manipulates the database directly. It can be in php or anything else you like. Take a look at the scripts in the "maintainance" directory in your mediawiki install. -- Finlay McWalter | Talk 23:16, August 29, 2005 (UTC)
Wikipedia:Bots is generally the place to be for bot-related discussions. The mailinglists serve their uses, but shouldn't be required to do anything. --fvw* 01:18, September 1, 2005 (UTC)

Can't load this diff

When I try to load this diff, Wikipedia fails and gives me a load of unprocessed wiki markup and the error Fatal error: Allowed memory size exhausted. Is it just me or is there a problem with Wikipedia (i.e. is anyone able to load it correctly)? --IByte 23:15, 29 August 2005 (UTC)[reply]

I get the following content, just if anyone gets anything different:
Line 159: Line 159: New Zealand's landscape has appeared in a number of television programmes and 
films. In particular, the television series [[Hercules: The Legendary 
Journeys|Hercules]] and Xena were filmed around Auckland, the film [[Heavenly 
Creatures]] in Christchurch. The television series The Tribe is set and filmed in New 
Zealand as well. Director Peter Jackson shot the epic The Lord of the Rings trilogy in various locations around 
the country, taking advantage of the spectacular and relatively unspoiled landscapes, and Mount Taranaki was used as a 
stand-in for Mount Fuji in The Last Samurai. Other movies currently filming in New Zealand include [[King Kong 
(2005 movie)|King Kong]] and The Chronicles of Narnia: The Lion, the Witch and the Wardrobe. New Zealand's 
landscape has appeared in a number of television programmes and [[List of New Zealand Feature 
Films|films]]. In particular, the television series Hercules and [[Xena: Warrior 
Princess|Xena]] were filmed around Auckland, the film Heavenly Creatures in [[Christchurch, New 
Zealand|Christchurch]]. The television series The Tribe is set and filmed in New Zealand as well. Director [[Peter 
Jackson]] shot the epic The Lord of the Rings trilogy in various locations around the country, taking advantage of 
the spectacular and relatively unspoiled landscapes, and Mount Taranaki was used as a stand-in for Mount Fuji 
in The Last Samurai. Other movies currently filming in New Zealand include King Kong 
and The Chronicles of Narnia: The Lion, the Witch and the Wardrobe. 

Fatal error: Allowed memory size of 52428800 bytes exhausted (tried to allocate 35 bytes) in /usr/local/apache/common-
local/php-1.5/includes/DifferenceEngine.php on line 1091

--Titoxd 23:22, 29 August 2005 (UTC)[reply]

That usually means that a vandal has overwhelmed the difference engine by instering lots of pasted material. --Golbez 23:25, August 29, 2005 (UTC)
That's what you get when you rely on technology invented in 1786. --R.Koot 01:03, 30 August 2005 (UTC)[reply]
Yes, I hope that naming coincidence was deliberate, not embarrassing. Superm401 | Talk 01:14, August 30, 2005 (UTC)
You can still see the revision itself: [1] looks like the vandal added a couple hundred screenfuls of "balah" to the page. --cesarb 02:19, 30 August 2005 (UTC)[reply]
Fixed -- Tim Starling 02:26, August 31, 2005 (UTC)

lowercase #redirect directives just broke

Why did the #redirect directive become case-sensitive? I've been typing them in lowercase for months, and today someone broke them all. —Michael Z. 2005-08-30 16:06 Z

They don't look broken to me, do you have an example? Note that redirects have the same case sensitivity as links, i.e. only the first letter is case-insensitive. For example, #redirect main page won't work. -- Tim Starling 02:32, August 31, 2005 (UTC)
I mean the word "REDIRECT" must now be in all-caps. If you type it in lowercase, Wikimedia treats the "#" as the first item in an ordered list. —Michael Z. 2005-09-2 21:09 Z
Doesn't seem broken to me, either. Could you link to a page where it's broken? —Cryptic (talk) 21:14, 2 September 2005 (UTC)[reply]
Well, look at that. It's been fixed. I was confused because previewing a page still makes it show up as a list item. Cheers. —Michael Z. 2005-09-2 22:03 Z

MS Publisher

I have written an article in MS PUblishers on WIndows XP. I want to switch to a Mac G4 - is there an equivalaent program for Mac where I can read what I wrote, not destroy the formatting and seeing the pictures? If this is the wrong forum please let me knwo where to post Thank you

Peter T Knoepfler tamas@u.washington.edu

you can get microsoft office for the mac that should be able to open it as probablly will openoffice but this isn't really the place to be asking such questions. Plugwash 19:58, 31 August 2005 (UTC)[reply]

user page help

I am trying to add myself in the Category:Wikipedians in the United States and the ohio sub category, with no success. Can someone help me? You can edit my user page, the codes are at the vary bottom of the page. --Admiral Roo 11:38, August 31, 2005 (UTC)

Never mind. I got it. --Admiral Roo 11:40, August 31, 2005 (UTC)

commas and dashes

There are a number of wikipedia articles where the title of the article uses a comma in place of a dash. For example, the television series Have Gun — Will Travel is listed in wikipedia as "Have Gun, Will Travel". Is there a reason to avoid using a dash in the title of a Wikipedia article? Rick Norwood 13:13, 31 August 2005 (UTC)[reply]

As of a few months ago, not any more. -- Curps 16:42, 31 August 2005 (UTC)[reply]

More efficient handling of indefinitely-blocked non-anon-IP user accounts

Very often a vandal sock user account is created, runs for a few minutes, and gets indefinitely blocked. 99 times out of 100, after realizing that it's been indef-blocked, that sock never even tries to come back, yet our block list gets cluttered up forever, taking up memory space and affecting performance.

Can we handle indefinitely blocked non-anonymous-IP user accounts a different way? Take them out of the block list and put them on a "banned-user login list" that only needs to be checked at login time, not with every edit. When a user on the "banned-user login list" logs in, an indefinite block is temporarily reinstated in the block list and the login proceeds, so they can read and edit their talk page, and so forth, while still being blocked. However, after 12 hours or so, the "banned user" gets logged out automatically and their block is removed from the block list (but they can log in again anytime). If setting a timer for each such logged-in "banned user" is impractical, we could just set a fixed time (once a day? once a week?) when every logged-in "banned user" get logged out, globally, and their blocks are purged from the block list.

The advantage is performance: nearly all of the time with nearly all indefinitely blocked user accounts, we only need to do checks at login time and only in relatively rare cases will we need to have them temporarily in the block list, where the checks get done with every edit.

At least this would work with user accounts. For indefinite anonymous IP blocks (open proxies, etc), there doesn't seem to be any way to avoid doing checks with every edit.

-- Curps 16:56, 31 August 2005 (UTC)[reply]

Even that is over the top, I seem to recall logins merely set a cookie that is a hash of the password, so users don't need to log in via the server if they have sufficient technical expertise. A more obvious approach would be to allow admins to set the password of an account to something that disallows logins altogether, with the appropriate loging (as a block). This could even be done transparantly, by storing appending a magic token to hashed password in the password column of the user table, which could be restored on unblocking.
Technical point, I believe cookie tokens are now randomized and unrelated to your password. If I recall correctly, this was changed after someone's password was found by brute force from a cookie. Dragons flight
Block lookups shouldn't be that big of a deal though, they can be done O(1) or O(log n), so unless the block lookup still shows up in a big way in the profiling, I'm not sure this is as much a technical issue as a user interface issue (the block list is a pain to load, which could be easily solved by not displaying all username blocks older than a month by default). --fvw* 23:39, August 31, 2005 (UTC)
I agree, while blocks could be optimized it would be a pain to implement for little gain. By the way, in case no one has noticed, the block list is now paginated and searchable so it is much faster and nicer to deal with. Dragons flight 01:26, September 1, 2005 (UTC)
So it is, and there was much rejoicing. --fvw* 01:34, September 1, 2005 (UTC)

How is the block list implemented? Is it in memory, or does it require a database lookup each time? Every single day brings a few more indefinitely blocked socks/vandals, and this will grow without end forever into the future. Can the current implementation handle that? -- Curps 04:57, 1 September 2005 (UTC)[reply]

The blocklist is cached so no database lookup is necessary most of the time. I believe the cache is only purged when the blocklist changes. The key aspect is a match of user id numbers, which could logically be implemented as a quick search or similar (i.e. O(log n) time), though I don't know if this is presently the case. The caching and opportunity for good time complexity suggest to me that it could be very robust to expansion. Dragons flight 06:59, September 1, 2005 (UTC)
Where are you all getting this from? I optimised the block overhead on page saves a week before this discussion began, it now accounts for only 3% of save time. And I optimised Special:Ipblocklist, and Brion implemented paging and searching, so there's no problem there anymore. It's O(N) and there's no caching other than the MySQL query cache, but that doesn't make it a good optimisation target. I'll put up some current save profiling data on m:Profiling. -- Tim Starling 15:05, September 1, 2005 (UTC)
My apologies. I guess I conflated some knowledge about what BlockCache was supposed to do with the rest of blocking. I realize now that this isn't even used for user blocking. Even running as a SQL call though shouldn't it be better than O(N) in time complexity (not that it is neccesary implemented that way), but searching for a specific user id in a list of user ids (and other data) is the kind of thing that really ought not to require looking at every row, if there was any concern that it was slow (does SQL not optimize that kind of thing natively?). I realize we may also be talking past each other. My reading of Curps comment was with respect to how block check is implemented on user actions not how the Special:Ipblocklist works. Dragons flight 15:23, September 1, 2005 (UTC)
The BlockCache class is still used, it's just that the actual caching part was commented out to fix a bug (much to my annoyance). So for a while, it was just loading the whole ipblocks table on every save, then discarding it. The reason we needed a block cache in the first place was for efficient handling of range blocks, which needed to be stored in memory in a particular format, different to the current schema. The optimisation I did recently was to have BlockCache load only range blocks, and to do a separate indexed query for IP and username blocks. The query to load only range blocks is currently not indexed, requiring a full table scan, hence the O(N). The next task for optimisation would be to change the schema to handle range blocks properly, and thus do away with BlockCache. It's probably not practical to cache block information in memcached, due to its poor reliability. -- Tim Starling 00:49, September 2, 2005 (UTC)

Adding other users sigs

The only stupid question is one that isn't asked. How do other users go back and insert signatures for previous users? SchmuckyTheCat 19:35, 31 August 2005 (UTC)[reply]

Like {{unsigned}}? [[smoddy]] 19:57, 31 August 2005 (UTC)[reply]
Or you can just do it manually. It's not that much more work, since the hard part is manually cutting-and-pasting the name and time in any case. Although, I suppose it's important to put something that indicates that the signiture was added later and not by the original user, so I suppose {{unsigned}} is a good standard way of doing that.Aquillion 01:19, 2 September 2005 (UTC)[reply]

question about backups

Reading some recent posts on village pump (proposals) got me thinking. What procedures does WP have in place to recover from a serious attack, should one occur? The last available dump of EN at http://download.wikimedia.org/ is about five weeks old. Is this really the last available snapshot, or are they taken more often? Would it be technically difficult to roll back to a particular point in time if necessary? Dmharvey File:User dmharvey sig.png Talk 20:04, 31 August 2005 (UTC)[reply]

Each edit has a timestamp associated with it, so you can always roll back to a certain point in time, even without backups. With a little coding you could even hack up something that could have each edit after a certain timestamp only be commited after being approved. --fvw* 01:16, September 1, 2005 (UTC)
Well, a little coding. It's technically feasible anyway. --fvw* 01:16, September 1, 2005 (UTC)

Let's not get away from the question: Do we pull regular backups onto tape or CD-foo media and store them offline? Preferably, offsite in a fireproof safe. Or are we just winging it? If the next hurricane hits Florida, will we have to go around, hat in hand, to our mirrors to try to reconstruct the database?

I'd really like to ask that the Community allow this question to stand until answered by a member of the development team. Then, we can throw peanuts. — Xiong熊talk* 19:12, 2005 September 2 (UTC)

Brion has been actively working on a script that will produce dumps more often and more reliably. So long as this happens, it's not really an issue of doing nightly tape backups because a lot of people download those dumps, effectively making backups all over the world. Incremental dumps have also been discussed, and now that we have xml dumps that wish may become a reality. If you are asking if Hurricane Katrina had hit Saint Petersburg, Florida instead of New Orleans, then yes we might have a problem with losing some data. Keep in mind though that we also have data centers in at least Amsterdam and France, with the Yahoo servers coming in South Korea. You can read the September archives for more information. Nothing I said should be construed as true or official :) --Alterego 19:28, September 2, 2005 (UTC)

Special:Movepage should refer to meta:Help:Renaming (moving) a page

It would be nice if the instructions that appear when you click on the "move" tab at the top of a page (taking you to Special:Movepage/PAGENAME) contained a link to meta:Help:Renaming (moving) a page. Possible wording:

...
WARNING! This can be a drastic and unexpected change for a popular page; please be sure you understand the consequences of this before proceeding.
Please read meta:Help:Renaming (moving) a page for more detailed instructions.
(form here)

I don't know if these instructions can be changed by sysops, or if it takes a developer, or what, so I am posting here rather than filing a bug report.

Zack 20:56, 31 August 2005 (UTC)[reply]

Try MediaWiki:Movepagetext (protected). Bovlb 21:00:02, 2005-08-31 (UTC)
I've added the link as suggested at MediaWiki:Movepagetext. Messages such as this are contained in the MediaWiki: namespace, you can see all of them at Special:Allmessages. Almost all the messages are protected, but can be edited by admins. I've copied this discussion to MediaWiki talk:Movepagetext for reference, but for future reference the village pump is an apropriate place to propose changes such as this as the MediaWiki pages are not heavily watched. Thryduulf 21:06, 31 August 2005 (UTC)[reply]
And somewhere, Willy on Wheels is seeing the new warning and slowly realizing why he keeps getting blocked. Aquillion 01:21, 2 September 2005 (UTC)[reply]

Account creation throttle

I've just found MediaWiki:Acct creation throttle hit which states that "Sorry, you have already created $1 accounts. You can't make any more.". Does anyone know what that limit is? Thryduulf 21:20, 31 August 2005 (UTC)[reply]

I don't have access to the mediawiki boxes, but the default is 0 (disabled). That may just be to avoid dependence on memcached (which it requires) though, if you want an authorative answer you should probably go to wikitech-l. --fvw<;FONT COLOR="green">* 01:11, September 1, 2005 (UTC)
10 per day -- Tim Starling 13:24, September 1, 2005 (UTC)
Do you think cutting it down might slow down Willy? ~~ N (t/c) 01:22, 2 September 2005 (UTC)[reply]
No, I think it would just annoy people using transparent ISP proxies. -- Tim Starling 02:02, September 2, 2005 (UTC)

MySQL Binary values

I was wondering how I can get php to insert binary values into the database for the wiki. If anybody has the code for this, please let me know. Thank you, shardsofmetal 00:03, September 1, 2005 (UTC)

Diff timeout

Was that big server slowness we just had related to the vandal who was posting very large pages? It might be worth lowering the CPU limit on the diff operation a little then. --fvw* 07:10, September 1, 2005 (UTC)

Personal photo on user page, copyright?

If I want to add a picture of myself to my user page, what copyright selection should I choose? natsukigirl 23:10, 1 September 2005 (UTC)[reply]

This link should be useful: User:Vorash/Wikiusers gallery. Lots of examples. Sometimes they don't add any copyright notice at all, others {{fairuse}}, {{gfdl}} and {{pd-self}}. It depends on how you want your picture to be distributed (personally, I'd not include a photo at all). — Ambush Commander(Talk) 23:35, September 1, 2005 (UTC)
Thanks... well, I don't mind putting a picture... I just figured it might be fun... also, since I have a Japanese monkier, I wanted to have a picture showing that I'm caucasian (don't want to mislead anyone, lol). I suppose I'll just add a link to a picture instead. Thanks for the advice. -- Natsuki Girl\talk

Invalid titles

I've run a process to find pages in the database with invalid/inaccessible titles and rename them. The ones with invalid characters or that conflicted with existing normalized forms have had 'Broken/' prefixed to their names, and are listed at Special:Prefixindex/Broken/

These are mostly leftover from old software bugs since fixed, and a lot of them are redirects created by page moves to corrected titles and can now be discarded. --Brion 01:33, September 2, 2005 (UTC)

Great. I will check them and delete the ones that do not need to be saved. --cesarb 02:06, 2 September 2005 (UTC)[reply]
I cleaned the first page. Will continue later. --cesarb 04:00, 2 September 2005 (UTC)[reply]
All done. --cesarb 22:04, 2 September 2005 (UTC)[reply]

CSS style for no printing?

I belive there is a CSS style that allows e.g. a paragraph or part of a table to show on screen but not when printed. Which style(s) would that be? With such styles, is it possible to discriminate between colour printers and black/white printers? --Eddi (Talk) 09:39, 2 September 2005 (UTC)[reply]

All you need to do is find a style in the print stylesheet that has display:none and voila. It'd be a hackish solution though... — Ambush Commander(Talk) 12:30, September 2, 2005 (UTC)
Thanks, but where do I find the print stylesheet so that I can look for the right style? --Eddi (Talk) 13:58, 2 September 2005 (UTC)[reply]
It's on MediaWiki:Monobook.css, the class is "noprint". --cesarb 14:34, 2 September 2005 (UTC)[reply]
Works fine. Thank you very much. --Eddi (Talk) 15:45, 2 September 2005 (UTC)[reply]

Deleting an article not in user contributions?

Is there a reason, technical or otherwise, that when an admin deletes a page it does not appear in their user contributions, I think it would be fairly useful if it did show up, then we could see what an admin has been doing, we (admins) would be more accountable. Martin - The non-blue non-moose 22:23, 2 September 2005 (UTC)[reply]

Special:Log/Delete serves this purpose. At a technical level, deleted content is removed from the article database and so shows up in no one's contributions. This includes all past contributions as well, so your contrib list will never include deleted articles. Dragons flight 22:30, September 2, 2005 (UTC)
Whoops sorry, i never noticed you could filter it out to see an individuals deletions. Mind you it would still be nice to see it in the contribs, but it isnt that important. thnks Martin - The non-blue non-moose 22:43, 2 September 2005 (UTC)[reply]

Greek characters

Can I put Greek characters into an article using the "alternative language" set up from Windows? 

Αθήνα

which only needs one press for each letter and looks fine on my screen here

But the article on Athens for example has 

&# 913;&# 952;&# 942;&# 957;&# 945; (without the spaces)

which looks like html and which takes a lot longer to work out/type

I have RTFM but cannot find anything definitive

thanks, John

Saltmarsh 11:37, 3 September 2005 (UTC)[reply]

Yes, you can add them directly. Before the upgrade to MediaWiki 1.5, they were automatically converted to the HTML character escapes you see in the article; but you still could add them directly (the conversion was done by the server). What you see is a leftover from before the upgrade (you can change them to the displayed characters, if you wish). --cesarb 16:23, 3 September 2005 (UTC)[reply]
In fact, I think that changing them is recommended: it allows for a much more intuitive reading of characters when editing. — Ambush Commander(Talk) 16:41, September 3, 2005 (UTC)
Yes, but we already have a bot doing it, so there's no pressing need to do it by hand. --cesarb 18:28, 3 September 2005 (UTC)[reply]
User:Curpsbot-unicodify is currently working on Polish articles, but I ran it on Athens just now [2]. Note that the bot only converts character entities that are commonly printable in default fonts, thus a few character entities (&#7944; , &#8134; etc) in the "Greek Extended" Unicode range (U+1F00 to U+1FFF) are left uncoverted. -- Curps 19:00, 3 September 2005 (UTC)[reply]
Saltmarsh 06:35, 4 September 2005 (UTC) thank you all - John[reply]

Editing a page and the Insert Box causing saving errors

Several times I have been bitten by the Insert box at the bottom of the standard skin, edit screen. It is frequently very slow to form and so with a typical edit of

  1. Edit a section
  2. [show preview]
  3. Yes looks OK [Save Page]

If the save page is done before the complete skin has been returned the section being edited is saved on top of the whole page. Looking a page refreshes after a show preview, the major component of the page which seems to be the slowest to draw is the "insert box" by a large margin. As the number of characters in the insert box has increased, its rate of drawing seems to be getting slower and this is causing more page save errors.

When I had a dial up connection to my ISP I put it down to a slow line. But now I have a broadband connection and it is still happening so I guess it is server related.

It has happened to me a number of times and I have seen it happen to others. eg

I suggest that the Insert box is turned into a link to pop-up page, for anyone needs these characters. After all for most page edits these additional characters are not used. This would have three advantages it would:

  1. Reduce load on the server if it did not have to put out these characters for every page edit;
  2. Reduce load on the network;
  3. Reduce the likelihood of a section being saved as the complete page because the whole page is slow to download.
    which apart from being annoying causes an unnecessary load on the database servers to save the large change and then save a large reversal for the fix.

--Philip Baird Shearer 12:39, 3 September 2005 (UTC)[reply]

The key technical problem is that the hidden tokens identifying whether something is a section or not aren't loaded till near the very end, so it is possible on a slow page load for someone to hit save before the token saying what section is being editted is available (it defaults to full page editting) I will report that as a bug. Dragons flight 16:32, September 3, 2005 (UTC)
See bugzilla:1181. Dragons flight 17:43, September 3, 2005 (UTC)

1.6

Should Wikipedia:MW 1.6 bugs exist? Does it already? (I actually haven't noticed any difference at all from 1.5, but now that we're running on the bleeding edge we definitely should have this.) ~~ N (t/c) 01:08, 5 September 2005 (UTC)[reply]

Can't we just handle them here? The 1.5 page was useful because we got a huge lump of bugs in one go. Now that there's just going to be a steady influx of bugs (or not, fingers crossed), I think WP:VPT can handle it. --fvw* 01:13, September 5, 2005 (UTC)
Then maybe we should take off the "1.5 bugs" notice. ~~ N (t/c) 01:29, 5 September 2005 (UTC)[reply]
Makes sense, I've removed it. If anyone disagrees feel free to put it back. --fvw* 01:33, September 5, 2005 (UTC)
Just don't forget to report bugs at http://bugzilla.wikimedia.org/ so we have half a chance of keeping track of them! Remember we've got literally hundreds of wikis here, several dozen active. Not all of our developers are even on this particular wiki, whereas we get notification of bugzilla entries from several directions (including email and live IRC notices). --Brion 07:10, September 5, 2005 (UTC)

Errors in header for main page

When I am at http://www.wikipedia.org, the following is displayed at the top of the page: Warning: main(./lucene.php): failed to open stream: No such file or directory in /usr/local/apache/common-local/php-1.5/CommonSettings.php on line 712

Warning: main(): Failed opening './lucene.php' for inclusion (include_path='/usr/local/apache/common/php-1.5:/usr/local/apache/common/php-1.5/includes:/usr/local/apache/common/php-1.5/languages:/usr/local/apache/common/php-1.5/templates:/usr/local/apache/common/php-1.5/extensions/wikihiero') in /usr/local/apache/common-local/php-1.5/CommonSettings.php on line 712


I have tested this in both FF and ie, but since this is an error in the code for the page, it doesn't matter.

I tried looking around for a webmaster@wikipedia.org account, but didn't see one listed anywhere.

Below is a screen shot of the error message.

Image:wikipedia.home.page.errors.png

Works fine for me right now. Dragons flight 04:14, September 5, 2005 (UTC)


Template for rollback summary?

The admin rollback feature leaves an edit summary of "Reverted edits by user contributions page to last version by User. Is this an already-existing template that normal users could use for their rv summaries? Boojum 14:49, 5 September 2005 (UTC)[reply]

I don't think you can use templates in edit summaries. ~~ N (t/c) 16:16, 5 September 2005 (UTC)[reply]

page-move protection error doesn't display properly

When a non-admin user tries to move a page that is protected from moves, the error message doesn't come up properly. I've addressed this at MediaWiki talk:Protectedpage. --Ixfd64 08:36, 2005 September 6 (UTC)

  • It would also be nice if the tab showed whether it was fully protected or just protected from moves. - Mgm|(talk) 11:20, September 7, 2005 (UTC)

Need help with subcategories

I've created {{disambig-cleanup}}, which adds a page to Category:Disambiguation pages in need of cleanup. That category is in turn part of Category:Disambiguation, so adding an article to the first should also add it to the higher-level cat automatically, right? The problem is, it doesn't seem to work. Articles get added to Category:Disambiguation pages in need of cleanup, but they don't automatically show up in Category:Disambiguation. The workaround we came up with is to just directly add both cats to {{disambig-cleanup}}, but that's not really the right way to do it (at least as far as I understand how subcats are supposed to work). --RoySmith 12:57, 6 September 2005 (UTC)[reply]

I think you might misunderstand the way the categories and subcategories work. The articles you've added to Category:Disambiguation pages in need of cleanup are in Category:Disambiguation, but only insofar as Category:Disambiguation pages in need of cleanup itself is in Category:Disambiguation. You don't see that category on the first page of Category:Disambiguation simply because there wasn't a sort-key on it (I've added one). But rest assured, the articles are there - within a subcategory. Grutness...wha? 13:11, 6 September 2005 (UTC)[reply]

I see the problem. The template originally had only Category:Disambiguation pages in need of cleanup attached to it - Category:Disambiguation was added several hours later. Because of the way template assignment to categories works, any articles that were marked with this template prior to the change in the template will appear to be assigned to the extra category, but will need a null-edit in order for them to actually show up in the category's lists. I've performed a quick null-edit on the dozen or so articles in Category:Disambiguation pages in need of cleanup, and they should now all appear in the parent category. Grutness...wha? 00:38, 7 September 2005 (UTC)[reply]

Back-button editing weirdness

Up until recently, if I saved a file after editing, then noticed a slight glitch, I'd just click to back button to get back into the editer, make the change, and re-save. No problems. In the last week or so, though, each time this happens I find I'm in edit conflict with myself. What has changed in the last few days that would have caused this, and is there any way of fixing this apparent new bug? Grutness...wha? 07:54, 7 September 2005 (UTC)[reply]

Are these section edits or full-page edits?
I've disabled the conflict-ignoring for section edits as it has a very nasty habit of corrupting data when doing multiple section edits, and this seemed the lesser of various evils. --Brion 10:30, September 7, 2005 (UTC)

Postal Codes

Hello,

I was curious if there was a way to import the postal code table and insert it into an excel sheet. Or can it be purchased?

thank you

Phantom Redirect

Can somebody tell me how it is possible for Conestoga_wagon to redirect to the French Wikipedia's article on Islam without anything showing up in the page's code or history? More importantly, how do I fix it *when* I see it in the future. --Adamrush 15:58, 7 September 2005 (UTC)[reply]

Conestoga_wagon works fine for me, showing the english language article. A look at the history of the article shows no sign of any redirect having been there ever. -- Finlay McWalter | Talk 16:11, September 7, 2005 (UTC)
I just clicked on that link in your text and it gave me the same error. Fuck, I hope I don't have some virus. One other thing, I am using Firefox (no extentions) --Adamrush 16:23, 7 September 2005 (UTC)[reply]
It works for me, and I'm using firefox. Zoe 21:26, September 7, 2005 (UTC)

Editor watchlist

I was thinking how great it would be if there was a function that is cosmetically similar to the present special:watchlist, but rather than watching pages, it watches users. Everytime they make an edit they go to the top of your special:EditorWatchlist. It would be great for monitoring potential vandals and also for seeing when friends are online etc.

Probably too technically difficult to be worth implementing, but it would still be great. Martin - The non-blue non-moose 21:01, 7 September 2005 (UTC)[reply]

There is an IRC channel which tracks recent changes. I can't find it right now, but I'm sure someone has written a program that monitors it, and can do stuff like you want. -- Finlay McWalter | Talk 21:08, September 7, 2005 (UTC)
It's CryptoDerk's Vandal Fighter. -- Finlay McWalter | Talk 21:09, September 7, 2005 (UTC)
Downloaded and installed. damn this program is good! thanks Martin - The non-blue non-moose 21:25, 7 September 2005 (UTC)[reply]

Newuserlog

I've installed my extension that logs new user creations at Special:Log/newusers, it's neat for finding new users to welcome, usernames to ban and for catching vandals. —Ævar Arnfjörð Bjarmason 22:20:00, 2005-09-07 (UTC)

Neat. It would be handy if each log line had a "contribs" link too. -- Finlay McWalter | Talk 22:30, September 7, 2005 (UTC)
Retrieved from "https://en.wikipedia.org/w/index.php?title=Wikipedia:Village_pump_(technical)&oldid=22800989"