RFC bandwith, packet size and latency

[Henning Schulzrinne]

I've gathered some additional data on the RFC publication process and 
imported the data into EDAS (a publication management system that I 
wrote). The results are at 
http://www.cs.columbia.edu/~hgs/internet/rfc.shtml

The results differ slightly from the earlier ones since I was able to 
match additional RFCs by their I-D tag. However, the basic gist is the 
same and the trend has taken a large tick upwards in 2002: the average 
delay for RFCs published in 2002 was 2 years and 2 months. (The maximum 
was 5 years and 2 months!) As discussed in more detail on the web page, 
this estimate is likely low, since there seems to be an increasing 
tendency in some working groups to let a draft go through a number of 
draft-personal stages before becoming draft-ietf-wg. If the draft 
changed its title during that time, I cannot match it and thus only 
measure the time from draft-ietf.

As others have pointed out, it's amazing how painful gathering this data 
is. Since everything is in more-or-less random text format, with 
occasional changes, parsing is prone to errors. I did a few sanity 
checks, but this is a good project for a forensical statistician.

Also, in recent years, even the IMR is incomplete in its I-D list, so 
that I have no data on a number of -00 records. This explains some of 
missing data points (where the number of published RFCs is much larger 
than the number of measured RFCs).

It would be trivial for the RFC editor to gather the "first draft" time 
from the authors. (This would also be helpful for IPR issues.)

If somebody wants to play with the data, let me know and I can give them 
SQL access. (No, not Microsoft SQL...)

Henning