On Tue, Oct 11, 2005 at 11:41:29AM -0400, Robert Story wrote: > On Sun, 9 Oct 2005 13:12:24 +0200 Magnus wrote: > MF> I can of course add another member, f_create_from_tstring_new or some > such, > MF> last in the struct and call that one if and only if the old one is NULL. > MF> With this change everything except for the size of the struct is > preserved, > MF> It also sucks since this kind of stopgap measurement tend to become > MF> permanent, how do I go about deprecating the old field? > > Just add a comment by the old field indicating the new, preferred, field.
Ok, have done so. New patch uploaded. > MF> Another point is that it seems the normal way to specify IPX addresses is > MF> network.host.port so the network:host/port style is a net-snmp invention. > > I'm not familiar with ipx, so I don't know what 'network' is in this context. > The idea is that net-snmp uses 'transport:xxx' - ie everything after the > tranport is interpreted by the transport code. Yes, what i am talking about here is the xxx part, and how it should be handled, so to write a full ipx address you would write ipx:12345678:9ABCDEF01234/5678. > MF> A third point is that if you explicitly tells the agent that the default > MF> domain for a certain service X is D, and then enters an address that > MF> subsequently is treated as beeing in domain E, then that is rather > MF> surprising. > MF> > MF> FYI: Currently the implementation does make everything starting with / a > MF> unix address. > > I'm loathe to change the default behavior, because I'm sure that there is code > that assumes things about peers that start with '/'. I agree about it but will maintain that said peers are making preposterous assumptions. The whole point is to get away from having anything but the transport code parse the xxx, and in my book that includes the leading '/'. > MF> Oh, yes, maybe I should try to explain my current scheme better: > MF> > MF> 1) If there is a transport specifier in the address, use it > MF> 2) If there is a specifier in the default_domain parameter, use it unless > MF> the address starts with a '/', in that case use unix. > MF> 3) If there is an application domain, use it > MF> 4) If there is an address and it starts with '/', use unix > MF> 5) Use udp > > Looks reasonable.. have you looked to see how much code change would be needed > for this? Not for this change and nothing else no, but I do not think it would need that much. I'd say something like 150 lines of diff for changing only the selction process and ignoring the new lookup code. Would that patch be of interest separately? > I'd also like comments from some other folks.. Me too :-) /MF ------------------------------------------------------- This SF.Net email is sponsored by: Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl _______________________________________________ Net-snmp-coders mailing list Net-snmp-coders@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
