Rao Shoaib wrote: > James Carlson wrote: >> Sebastien Roy wrote: >> >>> | inet_cidr_ntop | Volatile | Undocumented | >>> | inet_cidr_ntop | Volatile | Undocumented | >>> | inet_nsap_addr | Volatile | Undocumented | >>> | inet_nsap_ntoa | Volatile | Undocumented | >>> | inet_cidr_pton | Volatile | Documented[5] | >>> | inet_neta | Volatile | Documented[5] | >>> >> >> Why are those last two treated the same as the others? They appear to >> be intended to be used as programming interfaces by ordinary >> applications, and seem to be useful interfaces. Why should they be >> marked as "Volatile?" Are they in fact likely to change in incompatible >> ways? >> > One goal of this project is to move closer to ISC source base and make > merging and updating easier.
That's assumed. > Since ISC does not guarantee that these > interfaces will not change in incompatible ways we do not want to make > that guarantee. Are there any interfaces for which they make that guarantee? If there aren't any, then I suspect that the wrong test is being applied. Where in the documentation is it made clear that these things are effectively unusable? >> Does Volatile really match with the upstream behavior and the downstream >> usage? Or is it just a replay of "external?" >> > I am not aware of any ISC interface classification scheme, so how was it > figured out which interfaces ISC will not change. In fact ISC has > changed some interfaces in incompatible ways and for this update we have > to implement workarounds to provide backward compatibility and we do not > want to keep doing that. I think that's a mistake. There's no free lunch here. Either someone (ISC or OpenSolaris) provides compatibility, or applications will simply break. As the latter is clearly an unacceptable result, it means that future projects should be told by the ARC that they must not use this software. If that's the case, then why bother integrating at all? Does ISC really break documented interfaces in this way? Do you have specifics to share? > Your assumption about volatile being a replay > on external is correct. In that case, this is an error. -- James Carlson 42.703N 71.076W <carlsonj at workingcode.com>
