>>>>> On Thu, 20 May 2010 09:40:06 +0200, Magnus Fromreide >>>>> <[email protected]> said:
MF> Now, I would love to hear some kind of comments on this. I need to go read the actual patch details, but the ideas look good in general :-) MF> generate-netsnmp_transport.diff: IMHO, these should be auto generated in the first place at run-time like the init_() list is (and how all the mib-module defines are). So this is the right goal, but probably not the best exact solution. If you do things like this at autoheader/autoconf time it isn't as clean and extensible as it would be if you let the end-user drop in new transport implementations without needing to rerun autoconf (eg, right now if they create new mib modules they don't need to run autoconf). MF> I do require that at least one transport is configured but that MF> is open for debate, there is no real reason to require that save MF> that the agent becomes utterly useless without any transport but MF> on the other hand e.g. the Alias transport is enough to get past MF> that test so maybe I should just skip it. I think everything becomes useless if there is no transports, so I think we should at least warn them they've done something stupid. Always, IMHO, think of the end-user and do what's right for them (and what's right for them in this case is an error since it makes no sense otherwise). If they're actually managing to figure out how to config with only the 'alias' transport then, um, that's not an exception I think we should test for since they're obviously, um, smart? (Just the 'callback' transport would likely be not very useful either, but I can actually see third-party code wanting just that at least). -- Wes Hardaker Please mail all replies to [email protected] ------------------------------------------------------------------------------ _______________________________________________ Net-snmp-coders mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
