Peter.Memishian at Sun.COM wrote: > > Further, I'm still uneasy about the object of the *-prop subcommands > > being links. This is not consistent with the existing dladm(1M) > > usage, where verb and type of object are concatenated to form the > > various subcommand names. With the proposed *-prop options, this > > would imply that the objects are properties are the objects, when the > > objects are really the links. > > > > To be consistent with the current dladm(1M) model, the property > > subcommands should be made specific to the type of objects they > > correspond to, e.g. setprop-link, getprop-link, showprop-link. This > > would be also compliant with the latest CLIP guidelines. > > Yes, this is true -- and we also discussed this issue internally. The > consensus was that setprop-link seemed too cumbersome, and that the net > result might be reduced usability (e.g., one can easily see users getting > confused about where to put the dashes or words -- is it set-prop-link, > setprop-link, set-link-prop[1], ...).
If you think that there are issues with the latest CLIP guidelines, they should be discussed with UIRB, and addressed as part of the recommendation. Otherwise there's a risk that every command line interface faced with the same problems will take a different path, leading to more inconsistencies across the board. > > Of course, more discussion on this issue is welcome. > > [1] Of course, for an experienced user set-link-prop makes no sense as per > your original point -- but seems completely reasonable for the vast > majority not familiar with the finer points of dladm's user interface. > Indeed, that vast majority is the set that I see preferring set-prop > to setprop-link. I'm not convinced. You have the risk of designing the interface into a corner, which could lead to even more inconsistencies down the road, e.g. defining properties that need to be applied to different types of objects. A user already has to understand that the dladm subcommands act on specific types of objects. Once this is understood, subcommands of the form setprop-link seem a "natural" fit. Nicolas. -- Nicolas Droux, Solaris Networking Sun Microsystems, Inc. http://blogs.sun.com/droux
