I didn’t want to implement an open HTTPs server for webhooks, even if they look sexy, but I also wanted to load the Telegram servers as little as possible, so given that the long polling doesn’t really get any delay if you arbitrarily increase the timeout*, I setted up a 300 seconds timeout for a 5 minutes refresh of the request.

My getUpdates URL is built like so:

"https://api.telegram.org/bot".$telegrambot."/getUpdates?timeout=".$timeout."&offset=".$offset

where $timeout is set at 300.

What I found out from my PHP app logs is this:

Time | UpdateID | HTTP Response | Total time
20:56:09 455417146 200 50.210294
20:56:59 455417146 200 50.205697
20:57:49 455417146 200 50.18303
20:58:39 455417146 200 50.191815
20:59:30 455417146 200 50.180151
21:00:20 455417146 200 50.178455
21:01:10 455417146 200 50.204421
21:02:00 455417146 200 50.197673
21:02:50 455417146 200 50.193216
21:03:41 455417146 200 50.205001
21:04:31 455417146 200 50.190687
21:05:21 455417146 200 50.178421
21:06:11 455417146 200 50.198388
21:07:01 455417146 200 50.191959
21:07:51 455417146 200 50.190216
21:08:42 455417146 200 50.193751
21:09:32 455417146 200 50.192767
21:10:22 455417146 200 50.189805
21:11:12 455417146 200 50.205147
21:12:02 455417146 200 50.191382
21:12:53 455417146 200 50.191173
21:13:43 455417146 200 50.201623
21:14:33 455417146 200 50.190784
21:15:23 455417146 200 50.189714
21:16:13 455417146 200 50.191169
21:17:04 455417146 200 50.193477
21:17:54 455417146 200 50.199469
21:18:44 455417146 200 50.210404
21:19:34 455417146 200 50.183195
21:20:24 455417146 200 50.178331
21:21:15 455417146 200 50.203791
21:22:05 455417146 200 50.203149
21:22:55 455417146 200 50.190459
21:23:45 455417146 200 50.19248
21:24:35 455417146 200 50.193081
21:25:26 455417146 200 50.180961
21:26:16 455417146 200 50.197347

No matter what timeout value you set in the URL, the server will always return after 50 seconds.

Why oh why?

* you can check on Wikipedia for a long polling description, anyway here’s the hang of it: you load a URL with a set timeout in the query, the webserver accepts the request but doesn’t send back any data until either some updates are found, or the timeout expires. So, for example, if the timeout is 60 seconds, either there’s an update in the next minute, after which the server immediately responds with the update information, or at the 60 seconds mark the server will respond with “no updates”, at which point you can send out an HTTP request again to that URL, and the cycle starts anew. Since there is no real delay between the updates and the server response, it would be sensible to set a high timeout, even 10 minutes, so, while still getting updates as soon as possible, you will load the server very little.