Version 4, changed by s3admin. 10/20/2006. Show version history
Performance: get off your tortoise and run
Tags: [cloud, search, optimization]
Well, I've been thinking (for a long while now) about how to deal with performance problems on big wikis. The JDC is a prime example: there are 982 pages with tags (or something like that).
Tonight it occurred to me to try non jot:search methods of pulling in the list of tags.
I've got a demo that will map tags --> # of occurrences for any wiki you have access to: GlobalTagCloudTest.
Getting the raw data turns out to be quite quick (most of the time): even from the JDC, you can generally get a raw list of tags in under 20ms.
The trouble is parsing those: from the JDC, that's still taking about 2s -- way too long. So optimization may be necessary.
And obviously you don't want to plug in your user/pass every time you load a page. So that needs solving.
But it's a start.
Getting a global tag cloud for an arbitrary wiki might have some other uses, too... anyone got good ideas?
By PhillipKast at 03/19/2006 08:25PM (edit) | Comments (1) | Post comment
Back to ProjectBlog
ScottMcmullan said, 03/21/2006:
The performance issue will soon go away... Until then this is an interesting approach. More dialog offline or in the forums...