James Carlson writes: > Darren J Moffat writes: > > Printing and NFS are much more interesting services than remote login > > shell services. > > Yep; agreed. Even httpd and samba may be better targets. > > I think the problem here may be a collision between wanting to "show > progress quickly" and doing something that's useful on real networks. > Getting printing and NFS on board in a useful way is likely to take > quite a bit longer. > > For my part, I have no real objection to seeing inetd enhanced to be > able to advertise services. It's at least a proof of concept that > lets us work through the various issues (such as on which networks to > advertise and how to represent in SMF).
The thing I'd like to work through based on the list of services we actually want to advertize is whether this belongs in inetd, with inetd specific properties. The answer to the latter is almost certainly 'no' -- if the operation is "advertize this service", and that operation is generic across a set of common restarters (e.g. inetd, startd, and even the Sun Cluster restarter), I'd hate to have the interface/ administrative operation be gratuitously different. I agree that inetd may be the easiest place to do a proof of concept implementation, but if we could get more value out of a startd implementation (httpd, specifically, may be a nice intersection of non-complex advertizing model and useful service to advertize) or better yet a restarter-generic implementation, I'm not at all convinced it would be significantly more coding work. I don't want to stop the project team from integrating bonjour/mDNS better with SMF -- I primarily want to make sure that the interfaces we advertize are suitably extensible for the services we *really* want to advertize, because it's such a good idea. :) Oh, and to refer to your NFS-on-board comment... with Doug's recent move of NFS configuration into SMF, I'll obliquely note that I think the NFS participation could be driven out of SMF too. But, that's just a first-pass analysis. More thought would obviously be necessary. liane -- Liane Praza, Solaris Kernel Development liane.praza at sun.com - http://blogs.sun.com/lianep
