On Thu, Aug 11, 2016 at 5:59 PM, David Walker <dave@mudsite.com> wrote:
> Hi all,
>
> I'm looking to resolve this bug, and thought about 2 means in which to
> implement the request of exposing C's getaddrinfo() and the addrinfo
> structure. There's not really an equivalent means to accomplish this in
> userland without trial&error.
>
> So I'm not sure the communities best-practices on altering current userland
> function signatures to implement a new feature. So I've devised 2 means of
> implementing this feature but don't know which would be preferred. The
> former introduces new functions for addrinfo-specific usage. The second
> alters current socket_connect/socket_bind signatures. They are both very
> similar in their implementation. I'm filling out an RFC[1] for this
> addition, and have details of each implementation idea typed out in it for
> you to see what I'm trying.
>
> Suggestions and advice on best-practice greatly appreciated.
>
I feel the weight of new functions, but really getaddrinfo is meant to both
simplify and unify how sockets are dealt with in the modern era.
socket_getaddrinfo converts the returned linked list into an array of,
essentially, socket contexts. The user can inspect those contexts and
choose the one he wants to use in connect and bind. Simple.
With a few minor tweaks to naming (to lump all the addrinfo functions
together), this makes sense to me:
$contexts = socket_addrinfo_lookup($node, $service, $hints);
foreach ($contexts as $context) {
print_r(socket_addrinfo_explain($context));
}
$socket = socket_addrinfo_connect($contexts[1]);
// ...
So I would lean toward flavor #1. Thanks!
