[rrg] Name-based API, was: Re: presentation/discussion list

[Iljitsch van Beijnum]

On 24 nov 2008, at 1:51, Brian E Carpenter wrote:

>> I don't think the mere fact that applications must do the name to
>> address lookup and then hand the address to the network stack makes
>> renumbering harder.

> I think it does, for two related reasons.

> 1. The address lookup API does not (currently) return a lifetime,

Well, the application could simply perform an address lookup every  
time it needs the address. Most modern resolver libraries do caching  
so on those systems that wouldn't impact performance.

> 2. The fact that the upper layer has knowledge of an address with
> no known lifetime creates an irresistible temptation to store that
> address and use it again an undefined time later.

Well...

> We've known this was a problem for many years so maybe it's
> time to fix it, although that won't help with the legacy apps.

Right. A name based API does have some disadvantages apart from what  
you mention above:

- architecturally cleaner
- ability to automatically try multiple addresses at connect() time
- ability to use multiaddress multihoming through the likes of SCTP
- ability to work with new address families