Re: [PATCH 0/5] transport: unify ipv4 and ipv6 code paths

[Jonathan Nieder <jrnieder@gmail.com>]

Erik Faye-Lund wrote:
> On Thu, Mar 8, 2012 at 1:48 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:

>> Patch 4 is the heart of the series.  It provides an interface similar
>> to getaddrinfo that can be implemented on top of either gethostbyname
>> or getaddrinfo and puts each implementation in a separate file.  This
>> way, callers can just use the common API and they do not need to have
>> two copies of their code, one for each host resolution API.
[...]
> What happened to this series?

It seemed nice in principle, but if I understood correctly then the
API was too complicated for anyone but me to like it.  Instead of an
interface similar to getaddrinfo, it would probably be better to
implement an interface _identical_ to getaddrinfo for the the subset
of functionality that git uses.

See [1] for example for proof that that's possible.

I would make some different choices from Russ if doing it myself ---
e.g., I would want to unconditionally use

	#define getaddrinfo git_getaddrinfo

when NO_IPV6 is set so testing these code paths would be as simple as
enabling NO_IPV6, even on systems that also provide the modern API in
libc.

Hoping that clarifies,
Jonathan

[1] http://git.eyrie.org/?p=devel/rra-c-util.git;a=tree;f=portable