[ofa-general] Re: initial set of "direct" SDP tests in netperf

[Michael S. Tsirkin]

> Quoting Rick Jones <rick.jones2 at hp.com>:
> Subject: Re: [ofa-general] Re: initial set of "direct" SDP tests in netperf
> 
> >>However, I think that is the "better" problem - SDP is a "protocol" not 
> >>an address family.  (Again, based on what little I understand about 
> >>SDP).

True but it's not an IP-based protocol.
So IPPROTO_SDP is kind of weird: the comment in netinet/in.h
on my system says:
/* Standard well-defined IP protocols.  */

> >I see why this makes sens for you, but in what sense is it a "better" 
> >problem?
> 
> Because it isn't trying to describe a change in protocol as a change in 
> addressing.  It makes getting to SDP look like getting to any other 
> transport-layer protocol - eg UDP, TCP, or SCTP.  My intuitive guessing 
> suggests that fewer folks use getprotobyname() than getaddrinfo().

However, for people that do - protocol numbers are assigned by
IANA, while AF numbers are basically free-running numbers.

Thus using AF rather than IPPROTO_ could have been a way to bypass
the need for standardization.

-- 
MST