I just read this mini-essay here:
http://en.wikipedia.org/w/index.php?title=Wikipedia_talk:WikiProject_Chemist...
It is about the effect on the parser of large navboxes (and of other templates). It is a bit technical, but I thought it might be of interest.
And I don't know how I ever missed this:
http://en.wikipedia.org/wiki/Wikipedia:Overlink_crisis
It really is worth reading all that essay and its talk page.
Carcharoth
On 17/02/11 10:55, Carcharoth wrote:
I just read this mini-essay here:
http://en.wikipedia.org/w/index.php?title=Wikipedia_talk:WikiProject_Chemist...
It is about the effect on the parser of large navboxes (and of other templates). It is a bit technical, but I thought it might be of interest.
And I don't know how I ever missed this:
If navboxes could be loaded dynamically, with AJAX, then editors could put as many links in them as they liked without DoSing the servers.
That's not to say having huge category listings in collapsible boxes at the end of an article is a good idea, in terms of style and usability. But it would be nice to separate the performance issue from the style issue.
-- Tim Starling
On Wed, Feb 16, 2011 at 6:55 PM, Carcharoth carcharothwp@googlemail.com wrote:
And I don't know how I ever missed this:
http://en.wikipedia.org/wiki/Wikipedia:Overlink_crisis
It really is worth reading all that essay and its talk page.
Make sure you read the history too, where I removed all the bogus technical claims three different times over the span of a few months before giving up when Wikid77 re-added them a fourth time. All the claims that links per se are a significant technical problem are complete garbage. The links tables are large, but not any kind of bottleneck. The thing that makes parsing slow is massive amounts of wikitext, whether it's links or anything else. Deeply nested templates might be a particular performance issue, for instance, but links table storage is definitely not. But I've given up on trying to keep Wikid77's persistently uninformed arguments off that page.
On Wed, Feb 16, 2011 at 10:03 PM, Tim Starling tstarling@wikimedia.org wrote:
If navboxes could be loaded dynamically, with AJAX, then editors could put as many links in them as they liked without DoSing the servers.
AJAX would lead to a poorer user experience, since they'd take some time to open. Wouldn't it make much more sense to have some kind of HTML-level cache for templates? Normally you can't do that, because wikitext constructs might start and end in different templates. But perhaps templates could opt in to an HTML cache through a magic word to get improved performance, at the expense of behaving somewhat differently. This should work for the giant static navigation templates as long as they don't vary per-article in important ways. Although of course, you'd need someone who understands the parser to write the feature, while AJAX could be done by any Wikipedia editor who knows some JavaScript.
(This is clearly more of a wikitech-l discussion, though.)
On Thu, Feb 17, 2011 at 11:58 PM, Aryeh Gregor Simetrical+wikilist@gmail.com wrote:
On Wed, Feb 16, 2011 at 6:55 PM, Carcharoth carcharothwp@googlemail.com wrote:
And I don't know how I ever missed this:
http://en.wikipedia.org/wiki/Wikipedia:Overlink_crisis
It really is worth reading all that essay and its talk page.
Make sure you read the history too, where I removed all the bogus technical claims three different times over the span of a few months before giving up when Wikid77 re-added them a fourth time.
Ah, OK. I see. That essay should be userfied then, and I need to retract what I said on his talk page.
The point made on the talk page here, though:
http://en.wikipedia.org/wiki/Wikipedia_talk:Overlink_crisis#Implications_for...
...is exactly the point I've made in the past. Large navboxes transcluded over many articles swamps 'what links here'. I did see the parallel discussion on wikitech-l when David Gerard raised this subject there, and am hoping that something is done about it when someone finds the time. :-)
Carcharoth