I got the same response from Alex awhile back (and I think confirmed by
Doug).  And I'm seeing the same results, as well.  I'm pretty sure it's
20,000 per IP without regard to user.
Jesse

On Thu, Aug 6, 2009 at 1:24 PM, Dewald Pretorius <dpr...@gmail.com> wrote:

>
> Just some background. I talked with Doug about this a few months ago,
> because I observed in the Rate Limit Header of get calls that the
> 20,000 number decremented by user, not by IP address in aggregate.
>
> Doug informed me that he was going to hand the issue over to Matt, who
> was on vacation at that point, to look into when he got back from
> vacation.
>
> Doug specifically said that the intended behavior was for the 20,000
> rate limit to be by IP address only.
>
> So, the point I'm trying to make is, at one point the API did count
> the 20,000 rate limit per IP address per user, but that was a bug that
> should have been fixed.
>
> I have not checked whether it is actually fixed. But, it's easy to
> check. Just do a GET call from a whitelisted IP with one user's
> credentials, check the remaining rate limit number, and then do the
> same call with another user's credentials. If each call gives you
> 19,999 remaining, then you know the bug still exists, and consequently
> no IP rate limiting is currently being done.
>
> Dewald
>
> On Aug 6, 2:04 pm, Chad Etzel <c...@twitter.com> wrote:
> > Hi Dewald,
> >
> > I asked "The Powers That Be" about it, and that was the response I
> > got. However, I am double and triple checking because that does sound
> > too good to be true :)
> >
> > -Chad
> >
> > On Thu, Aug 6, 2009 at 1:01 PM, Dewald Pretorius<dpr...@gmail.com>
> wrote:
> >
> > > Chad,
> >
> > > Are you 100% sure of that?
> >
> > > I mean, in terms of rate limiting that simply does not make sense.
> >
> > > For my site, TweetLater.com, it would mean I have an effective hourly
> > > rate limit, per IP address, of 2 BILLION IP GET calls per hour!
> > > (20,000 per user for 100,000 users).
> >
> > > It sounds wrong to me.
> >
> > > Dewald
> >
> > > On Aug 6, 1:37 pm, Chad Etzel <c...@twitter.com> wrote:
> > >> Hi Inspector Gadget, er... Bob,
> >
> > >> Yes, the current whitelisted IP rate-limit allows 20k calls per hour
> > >> *per user* on Basic Auth or OAuth or a combination thereof.
> >
> > >> Go, go gadget data!
> >
> > >> -Chad
> > >> Twitter Platform Support
> >
> > >> On Thu, Aug 6, 2009 at 12:13 PM, Robert Fishel<bobfis...@gmail.com>
> wrote:
> >
> > >> > Well it seems as though Twitter is saying that 20k calls per user is
> > >> > the intended functionality. Chad or someone else can you confirm
> this?
> >
> > >> > Also if the correct functionality is 20k per ip per hour will you
> then
> > >> > fail over to 150 per user per hour or is it cut off?
> >
> > >> > Thanks
> >
> > >> > -Bob
> >
> > >> > On Thu, Aug 6, 2009 at 7:54 AM, Dewald Pretorius<dpr...@gmail.com>
> wrote:
> >
> > >> >> Bob,
> >
> > >> >> Don't base your app on the assumption that it is 20,000 calls per
> hour
> > >> >> per user.
> >
> > >> >> You get 20,000 GET calls per whitelisted IP address, period. It
> does
> > >> >> not matter if you use those calls for one Twitter account or 10,000
> > >> >> Twitter accounts.
> >
> > >> >> If the API is currently behaving differently, then it is a bug.
> >
> > >> >> I have had discussions with Twitter engineers about this, and the
> > >> >> intended behavior is an aggregate 20,000 calls per whitelisted IP
> > >> >> address as I mentioned above.
> >
> > >> >> Dewald
> >
> > >> >> On Aug 6, 4:09 am, Robert Fishel <bobfis...@gmail.com> wrote:
> > >> >>> Wowzers (bonus points for getting the reference)
> >
> > >> >>> It appears as if each user does get 20k (according to the linked
> > >> >>> threads) this is I think what they intended and makes apps a LOT
> > >> >>> easier to develop as you can now do rate limiting (ie caching and
> > >> >>> sleeping etc...) based on each user and not on an entire server
> pool,
> > >> >>> makes sessions much cleaner.
> >
> > >> >>> I am whitelisted and I'll test this tomorrow evening to make
> double
> > >> >>> sure but this sounds great!.
> >
> > >> >>> Thanks
> >
> > >> >>> -Bob
> >
> > >> >>> On Thu, Aug 6, 2009 at 2:53 AM, srikanth
> >
> > >> >>> reddy<srikanth.yara...@gmail.com> wrote:
> > >> >>> > With a whitelisted IP you can make 20k auth calls per hour for
> each user.
> > >> >>> > Once you reach this limit for a user you cannot make  any auth
> calls from
> > >> >>> > that IP in that duration. But the user can still use his 150
> limit from
> > >> >>> > other apps.
> >
> > >> >>> >
> http://groups.google.com/group/twitter-development-talk/browse_thread...
> >
> > >> >>> > On Thu, Aug 6, 2009 at 7:50 AM, Bob Fishel <
> b...@bobforthejob.com> wrote:
> >
> > >> >>> >> From the Rate Limiting documentation:
> >
> > >> >>> >> "IP whitelisting takes precedence to account rate limits. GET
> requests
> > >> >>> >> from a whitelisted IP address made on a user's behalf will be
> deducted
> > >> >>> >> from the whitelisted IP's limit, not the users. Therefore,
> IP-based
> > >> >>> >> whitelisting is a best practice for applications that request
> many
> > >> >>> >> users' data."
> >
> > >> >>> >> Say for example I wanted to simply replicate the twitter
> website. One
> > >> >>> >> page per user that just monitors for new statuses with
> authenticated
> > >> >>> >> (to catch protected users) calls to
> > >> >>> >>http://twitter.com/statuses/friends_timeline.json
> >
> > >> >>> >> Say I was very popular and had 20k people on the site. Would
> this
> > >> >>> >> limit me to 1 call per minute per user or would it fall over to
> the
> > >> >>> >> user limit of 150 an hour once I hit my 20k? If so how can I
> tell it
> > >> >>> >> has fallen over besides for simply keeping track of the number
> of
> > >> >>> >> calls per hour my server has made.
> >
> > >> >>> >> Thanks
> >
> > >> >>> >> -Bob
>

Reply via email to