Sergey Fomin

> Not mandatory but will help with efficiency. Just out of curiosity, how is the efficiency defined in this case? We can expose the `InitialConnWindowSize` and `InitialWindowSize` settings, which disable...

looks good to me.

A few comments that I mentioned during the call today: 1) arista reference does not describe the policy-based implementation with the NH match criteria 2) it does, however, describe the...

> There's a fundamental problem with the automation: it does not take into the account necessary ondatra/FNT/lab infra management connections. > > The `deny all` term will cause test failures...

I would argue that `port-speed` is simply not applicable to the LAG interfaces. re. >2. Create a new leaf specifically for aggregated interface speeds, of type uint64 this already exists...

Thanks Darren, (to continue the conversation from #3363) As an alternative to the proposed policer definition, I would suggest implementing something like `/qos/policer-policies/` (similar to `/qos/scheduler-policies`), where the ingress policers...

> introduced new OC subtree (/qos/input-policies) The operational use case is to configure schedulers (policers) on input subinterfaces Today OC requires definition of queues for schedulers but in practice, input...

Please stop using this deviation for this code snippet, this is not correct as I indicated in https://github.com/openconfig/featureprofiles/pull/2984#issuecomment-2108074887 Regarding this code snippet, I have an open issue #2977. If there's...

I believe `e.AutoNegotiate = ygot.Bool(false)` should also be under a deviation if there's a vendor that requires this config. Potentially, you can group all three things under a single deviation,...

>Divide ipv4 classifier type in 2 groups, group 1 classifies packets based on specific transport destination port + ipv4 source and group 2 classifies packets based on ipv4 source. This...