Re: [sipcore] I-D Action: draft-ietf-sipcore-sessiontimer-race-02.txt

[Paul Kyzivat <[e-mail]>]

On 9/26/18 3:05 PM, Christer Holmberg wrote:

>> Ideally we would rewrite the whole explanation and change the terminology. I guess this would be a bis.
> 
> There have been some discussion on whether we should do that - instead of just trying to fix pieces here and there.
> 
> But, as I have indicated previously, I first want us to agree on HOW to fix, and then we can discuss how to document it.

I'm starting to lose track of the things that have been considered, 
rejected, etc.

ISTM that we are still struggling with what happens in obscure cases. It 
seems that some cases will always break somebody when some parties 
implement the update and others don't. And who has to change to make the 
fix. This will not be entirely pain free.

It would be helpful to me to have some information about what 
implementations of timer exist in the wild, why it is used, and what 
limitations those implementations have.

For instance, I have a sense that a lot of UAs request timers on their 
own. I have no idea why. I would expect it to only be requested by 
proxies. And since a lot of intermediaries are actually b2buas rather 
than true proxies, the demand ought to be even less.

	Thanks,
	Paul