Hello,
I'm developping an opensource movie collection management software and I've chosen TMDB as default data source. For now, the API key has to be entered by the user in the application configuration file. I asked a few friends to test the application and they all groused about the need to sign-up on TMDB and to give information such as postal address or phone number to get the API key. I want my users to be happy, so I'm looking for an other solution.
My question is: do you have another solution for this problem?
If not, I'm thinking about implementing a kind of proxy between your API and the app. The proxy would be hosted by me and will export an HTTP api similar to yours (but not identical, because I can adapt it to the requests of my app, and it doesn't need to cover 100% of your API), with a little cache layer to avoid flooding your API. What do you think about this?
Thanks, Antoine.
Un film, une émission télévisée ou un artiste est introuvable ? Connectez-vous afin de créer une nouvelle fiche.
Vous souhaitez évaluer ou ajouter cet élément à une liste ?
Pas encore membre ?
Réponse de Travis Bell
le 19 février 2013 à 18h47
No. Not at this time.
That is going to be potentially problematic for you as we do API rate limiting based on IP address. With 1 server doing all of the requests to our API, it wouldn't take long before you starting hitting the rate limits.
Your best bet is definitely create a full cached layer between us and you only make external calls as you are filling in data. Once you have enough of the dataset cached the number of requests you make will get lower and lower.
Cheers.
Réponse de NaPs
le 20 février 2013 à 15h07
Thanks for your answer. Putting aside the technical problems, you are not opposed to the idea?
Réponse de Travis Bell
le 20 février 2013 à 18h48
The local caching idea? No, it's what we recommend doing. Thousands already do this, and is definitely recommended.
Réponse de NaPs
le 21 février 2013 à 04h44
Nope, I'm talking about the idea of hosting a public proxy to your service for my users :-). Here is an implementation of the proxy I'm talking about: https://github.com/NaPs/Kubrick/blob/master/kubrick/tmdb_proxy.py
Réponse de Travis Bell
le 21 février 2013 à 10h07
That's fine, not much of a difference to us either way. Wouldn't this still use the IP of your server (and not that of the end user) to connect to our APi though?
Réponse de NaPs
le 21 février 2013 à 11h43
Yes, my server will always do the queries to your API instead of the user.
Réponse de Travis Bell
le 21 février 2013 à 11h48
Yes, so as I mentioned above ^ that will only work so far. We do rate limiting based on IP address, not API key. With more and more users your single IP will be doing more and more requests and will very likely easily trip our rate limiting.
Réponse de NaPs
le 21 février 2013 à 11h53
Is it possible to get an higher limit for my ip? Else I will try to make my queries from various IPs...
Réponse de Travis Bell
le 21 février 2013 à 12h28
Unfortunately, no. We do not issue custom rate limits.
Réponse de NaPs
le 21 février 2013 à 13h43
Ok, thank you for your help. I will announce my software soon, we'll see if it works :-).
Réponse de Karl Dietz
le 24 février 2013 à 14h15
Might be an idea to add support for X-Forwarded-For so ISPs with web proxies, or this custom tmdb proxy, are not penalized when they actually take load of the main servers.