> => this is like playing with UIDs, which are in the context of the > application and can be set by set[e]uid() & co, to permit or deny access to > privileged ports. My idea is to tune the context before performing some > operations and to reset it after to its previous setup.
While the semantics of such an approach is well-documented and understood in the uid space, I think it is hard to introduce new such mechanisms due to the side effect dangers. Assume you introduce env/context source address selection knobs and that both getaddrinfo() and connect()/sendto() will use them. First problem that appears is that internal to getaddrinfo() sockets might be connected (e.g. to connect to a DNS server, LDAP, whatever else). It is hard to guage the possible impact of such side effects, but I think they in general make the system much harder to understand. (Yes, they are a form of global variables and I think passing explicit arguments results in less complexity that using global context.) Erik -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
