[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
> 
> >>I was poking around - it would be nice if they could take AF_INET_SDP - I 
> >>have to wonder if IPPROTO_SDP is actually better, but seeing there has 
> >>been some discussion there (but not having read all of it) I'm just going 
> >>to go with the flow...
> >
> >
> >Basically everyone said "it does not matter".
> >Do you think IPPROTO_SDP is better?
> 
> To the extent that I have no idea what is really happening under the 
> covers with SDP I would say yes.
> 
> My understanding is that the only difference is that "SDP" is used 
> rather than "TCP."  That being the case, then I would think it 
> would/should be like using say UDP vs TCP vs SCTP (ignoring the obvoius 
> protocol differences).
> 
> Each are "INET" sockets using "INET" addressing, the difference is the 
> layer-four (transport) protocol being used, which is selected via 
> IPPROTO_TCP vs IPPROTO_UDP vs IPPROTO_SCTP.
> 
> And when/if IPv6 is supported, then there shouldn't (?) be any need to 
> have an "extra" AF_INET6_SDP - one would use AF_INET and AF_INET6 with 
> IPPROTO_SDP.
> 
> Also, an application making use of getaddrinfo() (as all well-written 
> apps are supposed to be these days :) wouldn't have to worry about name 
> to IP resolution in the general case (where a protocol is not provided 
> with the hints) when wanting to use SDP directly - it still calls 
> getaddrinfo() with AF_INET, AF_INET6 or AF_UNSPEC as before, no need to 
> worry that AF_INET_SDP is not groked by getaddrinfo().


Surely too late for OFED 1.2, but we can reopen this afterwards.

Are there disadvantages to using protocol instead of family?
Would this change present a problem for people using e.g. getprotobyname
to get the protocol number?

-- 
MST