Using multiple Twitter applications in a single tool

rate-limits

#1

Hello community,

We are researchers experimenting with some new ways to understand Twitter graphs (e.g. follower networks). We are running an algorithm that uses the friends/followers list endpoints to look for overlaps between the friend/follower lists of sets of related users. Unfortunately, the current rate limits (30 app / 15 user x 200 per request) make this process quite slow as even a modest friend/follower count can consume all of a window’s quota.

The goal behind the algorithm is to support a rapid interactive process of identifying common relations. The only way we can see to conduct proper user evaluation is to use multiple applications (e.g. ~10) in a round-robin fashion to multiply the quota to a reasonable level. However, from reading the Twitter terms and this forum it’s not clear whether this is permitted or not. We did tentatively try using two applications simultaneously for a handful of requests and that seemed to work fine, but we’re not sure how the API might react to continuous use of multiple applications owned by the same account for significant periods of time. By significant I mean sessions of 10-15 minutes, not 24/7 activity.

I’d appreciate any thoughts on this, particularly from Twitter staff.


#2

Using multiple application keys to essentially circumvent the rate limits is always disallowed and you should 100% not do this. Your apps and any associated IP addresses may be subject to suspension without warning if you do so. Thank you.


#3

Hi Andy,

Thanks for your reply and clarification. I did think as much which is why I wanted to ask before trying. I’d like to point out, however, that this the algorithm is being deployed as part of an academic research study, not a mass market end-user app. If there is any scheme that we can work under that allows a larger rate limit for specific end-points I would appreciate knowing how to apply.


#4

You may want to look at gnip. Their pricing isn’t readily available and they encourage you to exhaust the streaming api’s limitations before you use them but they might work something out since it’s an academic research study.


#5

Thanks for that. Will look into it.