Hi!
I think you are wrong. POSIX does not say that it's wrong to have the computer name associated with 127.0.0.1 in /etc/hosts, and it doesn't say that gethostbyname(my_name) should return a "real" IP.
OK, but it also doesn't explicitly state that gethostbyname(my_name) may return 127.0.0.1, so it's just neutral :-).
If I remember correctly, though, Microsoft, specifies that gethostbyname(my_name) *should* return real IP.
I don't know M$ standards, sorry, cannot comment on this.
This subtle difference means that Wine must take extra care. Fetching a real IP on a UNIX host is not trivial, but doable.
Yes, it is.
BUT: man 5 hosts tells that:
"The Berkeley Internet Name Domain (BIND) Server implements the Internet name server for UNIX systems. It aug ments or replaces the /etc/hosts file or host name lookup, and frees a host from relying on /etc/hosts being up to date and complete."
So it seems to imply that the DNS info is currently an authoritative source for host name queries. Of course everybody knows it, I'm just making note of it to support my following question: Did you ever see in the real world that a DNS server would return 127.0.0.1 as a response to a regular host name query ? I think not. And on the other side, try to query a DNS server for "localhost". Every DNS server should immediately return 127.0.0.1, it's RFC compliance requirement. Isn't it logical to keep the host table coherent with DNS ? Or are you happy that in case your DNS server fails you get totally different results ?
Of course I cannot argue more. It's just about the administrator's personal style and knowledge. I'm maintaining and administering Unix systems from 1985, long before Linux was born, and I would never allow such a thing in my host table. But if the current distro maintainers are of another view, I will not do anything against it. It's just about Open Source freedom: I will remove that patch from wine for me, and that's it :-).
With regards, Pavel Troller