This is the sort of thing that came to mind around the recent Twilio stock price drop driven by Uber's decreased use.<p>The whole article is about mobile messaging, but SMS isn't involved at all. I understand Airbnb does use SMS, but apparently only for specific use cases.<p>It's a bid sad, I think, that SMS is being displaced by either in-app messaging services or brand aggregated things like FB messenger. SMS has the advantage that if I don't like my mobile network, I could change providers and not change my address. The whole messaging space is now set to be highly fragmented.
Nice read. This is something we wanted to solve for our mobile applications too but we don't get enough traffic (per article, you guys are ~100k msgs/hr) to justify the engineering time required for this.<p>I don't use Airbnb as often as I should but I'm wondering do you guys have data on the users who simply use text messaging on the mobile devices (or other instant messengers?).<p>What led to the decision to spend time optimizing mobile messaging within the Airbnb app than just finding an easier path to connect users through text messaging?
How is the latency when the app is open and waiting to receive a reply?<p>From one of the diagrams it sounds like the API requests a delta update if it receives a push notification, but if the user has disabled push notifications, will the client still know to request an update? Also receiving a notification and then using a new connection to request an update is still slower than something like a web socket/long poll connection where new updates are immediately pushed to clients. Would love to hear how this works.
I've only skimmed the article (as of yet), but how is this superior/better than just a dedicated server plus DB for messaging with a simple READ(all), REPLY(to_message) API?<p>This seems like an interesting technical problem, but I can't help but feel that there's a significantly easier solution available. As a bonus: other than the fact that you can see messages in the app, I wonder how this is better than simply using SMS or a personalized anonymous email address per host-guest pair.<p>I've seen variants of this particular problem solved so often I'm surprised there's no open source back-end/platform that solves the back-end piece of this.<p>disclaimer: I haven't worked with anything on this scale, personally.