Akamai Bad Request (400) Restful API Delete method

 TL; DR: If you don't have time to find the root cause, just change Delete method to Post.

 In the past several weeks, we were experiencing an issue with Akamai: in a special sequence like Get + Delete + Get, or Post + Delete + Post, the request after Delete returned a Bad Request with reference no (xxxx.xxx.0) from Akamai edge server. But it never happened in our own environment, even the prod evn behind Akamai.

 We had some meetings with Akamai support engineers. They suspected that there may be some function keeping the HTTP connection with their edge server. At first, they thought our endless scrolling function caused the issue. But in fact we just load more data at the first time, then render only the first part of the data in the screen. When user scroll, we load other parts.

 After they checked their log, they found browser is using the same TCP connection for our Delete request. Which happens in the following case:
   Get /favourites  => setting up a connect with content-length some number
   Delete /favourites/{id} => using the same connection, with content-length xxx
   Get /favourites => bad request (400) from Akamai

In this case, Akamai edge server is sending Keep-Alive header to browser, so browser may use the same TCP connection to send the Delete request => Akamai edge server counts Delete request content-length as something not correct => edge server is waiting for more data from the next request => so next Get doesn't count as a separate request, and because it doesn't match what edge server expects, edge just returns 400 bad request to browser.
(It happens as if something like nullPost, nullGet when content-length is not correct - by support engineers).

 Their explanation sounds reasonable, but I used Fiddler and Wire Shark to check the content-length for all the requests,  couldn't find any incorrect data. All numbers seemed to be correct.

 We didn't have enough time to look into this issue further, and thinking of the request after POST never fails, I changed our Restful API Delete method to Post. This fixed everything.

 I am still not sure where the problem is, but at least changing Delete to Post makes everything good...

 Hope in the future, I can get some chance to make this clear. :)