Call-ID is recommended to use the form "localid@host" in RFC3261

[Jerrylove]

No description provided.

[sgodin]

Hi Jerry,

Thanks for the contribution! However, I'm not sure it could be worded as "recommended" by the RFC, unless I missed something. The only text in the RFC pertaining to this that I could find was:
Implementations MAY use the form "localid@host".
Some sample call id's in the RFC use that form, other's do not.

I'm not sure I see the value of merging this change, it will just make the messages a few bytes longer.

[adamroach]

On 10/26/18 10:47 AM, Scott Godin wrote: Hi Jerry, Thanks for the contribution! However, I'm not sure it could be worded as "recommended" by the RFC, unless I missed something. The only text in the RFC pertaining to this that I could find was: Implementations MAY use the form ***@***.***". Some sample call id's in the RFC use that form, other's do not. I'm not sure I see the value of merging this change, it will just make the messages a few bytes longer.

In practice, most products in the wild don't use the "@host" suffix. The only real requirement here is (statistically probable) uniqueness. The use of "@host" to ensure this is a relic of the time that 3261 -- or, really, 2543 -- was published (as it was thought at the time that the use of unique host names would allow for relatively short IDs). In practice, with so many hosts being NATted and therefore sharing their local notion of identity (e.g., there are a lot of "192.168.1.64"s on the Internet behind NATs), this approach has fallen out of favor. /a