Richard L. Hamilton wrote: > Getting rid of mipagent itself and its config files may be reasonable if > it is an orphan, and unusable as-is. > > However, if anyone might ever have use for Mobile IP (hard to say if it > was only usable in Solaris 9), ripping out the kernel hooks seems a bit > extreme to me, esp. the ones marked "stable". In particular, I wonder > if the IP_RECVSLLA option might have use beyond just Mobile IP implementation.
Regarding IP_RECVSLLA, I think I agree and that I hastily included it as part of the interfaces being removed. The reason it was included there was because mipagent is its only known consumer. I agree that it might have uses beyond Mobile IP. I'll move it to the "not removed" list. > Also, it seems to me that the demand for the infrastructure functionality > (home agent, foreign agent) isn't necessarily the same as the demand for > whatever functionality is required by a mobile node; the latter might > be quite desirable for laptops and such. (it's not clear to my by a very > cursory > reading of Mobile IP docs just what functionality a mobile node needs over > and above basic IP functionality) Theoretically, a mobile node implementation would not make use of the kernel infrastructure that's being ripped out, namely the strange source/interface based forwarding tables. I don't believe that this work would preclude a future project from implementing an IPv4 mobile node. > Getting rid of a nonfunctional daemon with no current customers, I > understand. Getting rid of the facilities which could support an > independent effort to get it up to date and functioning properly again, > that I'm not so clear on the advantage of. Removing dead and broken code alone will not erase it from existence. We have SCCS history and we keep deleted files, so the code will always be available to anyone who wishes to go back and pick it up again. In its current state, however, it's detrimental to our ongoing projects, so the short term plan needs to either be to remove it, or fix it. I have a problem with projects perpetually massaging the code to simply get it to compile knowing that it doesn't work and it's not of current use, just in case someone might want to play with it later on. The code itself will never cease to exist. It's just being removed from the onnv gate. -Seb
