In message <4EB85F14-0F65-45F6-8DF4-F11A8EE638FD@muada.com>, Iljitsch van Beijn um writes:
On 23-aug-2005, at 23:55, Steven M. Bellovin wrote:
This is exactly why people shouldn't implement drafts except possibly as a private in-house feasibility study.
In general, you're right; however, BGP documents have a special status. Because of how crucial BGP is to the Internet's functioning, I-Ds won't progress to RFC status (at least as Proposed Standard) without two interoperating implementations.
Ah, that makes sense. So how does that work for work on TCP (which is even more crucial than BGP)? You have to have interoperable implementations before writing the draft?
No. TCP is end-to-end; a problem shows up on that connection. By contrast, a BGP issue can affect everyone else, since your peers see only what you advertise based on your policy and what you've learned from others. Put another way, your problems (or your implementation's problems) affect others. That's not true for TCP, with the exception of congestion control behavior.
(I knew the IETF had some trouble with its internal organization. I had no idea it was this bad.)
Some would say that it's a feature -- rely on running code. --Steven M. Bellovin, http://www.cs.columbia.edu/~smb