It seems to me that this entire discussion is based on the premise that renumbering is extremely slow, difficult, expensive and disruptive. This belief is probably justified given current technology. Our Chemistry department reported that they renumbered 20 workstations, and it was a week before anybody involved got any useful work done. One shudders to think what this means if extrapolated to the global Internet. However, if we could develop a technology which could make renumbering of a network (ie. a company, university or similar-sized unit) fast and efficient, this premise would break down and it would no longer be unreasonable to consider renumbering the organization when switching providers. Suppose, for example that Domain Names were sacred but a daemon could somehow touch every machine in a certain domain simultaneously and give it a new number, while simultaneously updating the routing tables that pointed into that domain? No, I don't have an RFC written, sorry, it's just an idea. -- Walt -------
This belief is probably justified given current technology. Our Chemistry department reported that they renumbered 20 workstations, and it was a week before anybody involved got any useful work done. One shudders to think what this means if extrapolated to the global Internet.
The problem is that there are many things that can be easily overlooked. I believe if done right, renumbering 20 workstations in a single department is a no-brainer. However, what it would require is some cheat sheet of things that can be easily overlooked. Some recipe of how to do it right. I still believe renumbering what we reasonably can is well worth it and people should be encouraged to do so. But we *will* hit the kind of attitudes that Marty is pointing out, no doubt. Have seen my share of it already (e.g., net 35). I guess the question is whether we will try to make the best of it, expecting that while people prefer to be lazy, but even more prefer to be good citizens, whether we cut our losses and move to NSAPs or so, or whether we are just gonna let the network just go to hell. Hans-Werner
On Fri, 15 Apr 94 16:50:26 MDT Walt Haas wrote:
It seems to me that this entire discussion is based on the premise that renumbering is extremely slow, difficult, expensive and disruptive.
This belief is probably justified given current technology. Our Chemistry department reported that they renumbered 20 workstations, and it was a week before anybody involved got any useful work done. One shudders to think what this means if extrapolated to the global Internet.
However, if we could develop a technology which could make renumbering of a network (ie. a company, university or similar-sized unit) fast and efficient, this premise would break down and it would no longer be unreasonab le to consider renumbering the organization when switching providers.
I don't believe that renumbering should be this painful, *if done right*. In a previous occupation, I managed a network of about 150 machines of various types and kinds. It was intentionally set up in such a way that network configuration information (IP addresses, netmasks, default routes etc) was concentrated on one machine and other config info was built from that automagically (using machine-built config files, RARP, bootp, rdist and the like). I have done a number of splits, moves, merges and the like and found that, if the network is managed this way, renumbering doesn't have to be painful at all. The last big action (merging three parts that were split one year beforehand), took only two hours. The key to this is realizing that this might happen one day and planning for it. Basically, preparing new master config files and typing 'make' should do 90% of the work. It also does involve doing things *different* than most manufacturers want: instead of typing in the (fixed) IP address the first time a shiny new box boots, think how you can make (re-)configuration automatic. As an added bonus, you can use the same mechanism to distribute patches, new security measures and the like. I know of similar mechanisms at some other places. Like me, they have done some re-designs of their network over time. It can be done. Geert Jan
You mean assert network numbers like Appletalk allows for?!? On Fri, 15 Apr 1994, Walt Haas wrote:
However, if we could develop a technology which could make renumbering of a network (ie. a company, university or similar-sized unit) fast and efficient, this premise would break down and it would no longer be unreasonable to consider renumbering the organization when switching providers.
Suppose, for example that Domain Names were sacred but a daemon could somehow touch every machine in a certain domain simultaneously and give it a new number, while simultaneously updating the routing tables that pointed into that domain?
No, I don't have an RFC written, sorry, it's just an idea.
/***********************************************************/ /* Jon 'Iain' Boone Production Engineer boone@prep.net */ /* (412) 268-7874 PREPnet iain+@cmu.edu */ /* The opinions expressed above are just that: opinions. */ /* They belong to me and me alone. */ /***********************************************************/
participants (4)
-
Geert Jan de Groot -
hwb＠upeksa.sdsc.edu -
Jon 'Iain' Boone -
Walt Haas