On Mon, Jun 16, 2008 at 01:28:34PM -0700, Garrett D'Amore wrote: > It sounds, more and more, like the right answer is to fix the brain > damaged script. Yeah, it creates some burden for testing, but all the > other courses of action sound like they are distinctly inferior.
If we want to preserve canonical service names assigned by customers, then yes, I agree. For example, using i.services to merge the services file in ONNV with the IANA file shows results in: % grep 2049/tcp result nfsd 2049/tcp nfs # NFS server daemon (cots) shilp 2049/tcp nfs 2049/tcp % This shouldn't result in incorrect behavior. But, it's certainly not correct either. Instead it should resut in: nfsd 2049/tcp nfs shilp (I'm not sure the comment is useful. Certainly merging comments won't be easy, or even feasible.) > I'm rather severely opposed to breakup of /etc/services. At a minimum, > the additional ARC overhead for such would probably far outstrip the > cost of additional testing. I think fixing i.services is just a bugfix, but it might be good to file a self-approved ARC case since i.services's behavior would be changing in a somewhat significant way. Nico -- _______________________________________________ networking-discuss mailing list [email protected]
