Re: [regext] EPP and rate-limiting

["Francisco Obispo" <[e-mail]>]

On 17 Jan 2020, at 7:21, Patrick Mevzek wrote:

> If you look at what happens in the wild, you can come to the conclusion that there is probably no proper way to handle it in EPP... as many registries with this need do things in very different ways:

Hi *,

I agree with Patrick here, there are other tools the RY can use to handle situations like this.

In our EPP implementation we have some *reasonable limits* configured to deal with this type of situation, if the client is triggering a >=2000 error code and will continue to send the command, we basically delay the response as Patrick mentioned.

We’ve had a few interesting situations that we’ve observed. We had a registrar that was sending thousands of (epp `<hello />`) commands per second, this wasn’t affecting us as the `<hello>` processing doesn’t require a lot of work, but once we detected it,  a simple conversation with them was able to detect and fix a problem with their client implementation.

We have another situation where the client is sending a broken EPP `hello` command, triggering a syntax error, but that itself resets the timeout ticker, so they never notice it. We’ve contacted them about it and hopefully they’ll fix it.

Sending an *error* code requires the client to implement some sort of rate-limiting awareness making it non-trivial, whereas a delayed server response is handled automatically in most cases.

Best,



Francisco Obispo
VP - Technology
--------------------
Uniregistry

400 Spectrum Center Dr. Office 1660
Irvine, CA 92618

Office: [phone] x4202
[e-mail]