tfinc at wikimedia
Mar 15, 2012, 11:10 AM
Post #5 of 6
On Wed, Mar 14, 2012 at 9:30 PM, Tim Starling <tstarling [at] wikimedia> wrote:
> On 15/03/12 15:17, MZMcBride wrote:
>> I occasionally get asked about what a reasonable rate for querying the API
>> of a Wikimedia wiki is for a particular script or tool. I don't really know
>> the answer other than "be reasonable" and "specify an informative
>> User-Agent." That's essentially what I said when asked here:
>> If there's an authoritative answer (on Meta-Wiki or mediawiki.org or even
>> wikitech), that'd obviously be ideal. I'd also settle for a mailing list
>> post. I looked at <https://www.mediawiki.org/wiki/API:Main_page> to see if
>> it mentioned "limit" or "rate", but nothing came up. It's a fairly common
>> question; it should probably be a bit easier to find.
> When people ask me that question, I tell them to limit the
> concurrency, not the request rate. You can't do much damage with a
> single thread, whether queries complete in 10ms or 10s. But you can
> certainly do a lot of damage if you send a query once per second that
> takes 100 seconds to complete, using 100 concurrent clients.
> This is especially relevant for people who write toolserver scripts
> and the like. It's easy to write a server-side script which
> accidentally allows 100 concurrent connections to Wikimedia, when 100
> people happen to use it at once, or if someone decides to try a DoS
> attack using the toolserver as a proxy.
> -- Tim Starling
> Wikitech-l mailing list
> Wikitech-l [at] lists
We already have an article on this :
Looks like it captures Tim's point about slamming us with concurrency.
For those that regularly get asked about this ... what's missing from
it and how can we make it more discoverable? Currently its linked from
the front of the API page.
I changed the text slightly from 'Etiquette' to 'Etiquette & Usage
Limits' so that its a quicker find. Updates as needed.
Wikitech-l mailing list
Wikitech-l [at] lists