mcb@ncis.tis.llnl.gov (Michael C. Berch) (04/30/89)
In article <68215@pyramid.pyramid.com> csg@pyramid.pyramid.com (Carl S. Gutekunst) writes: > In <163@ncis.tis.llnl.gov> mcb@ncis.tis.llnl.gov (Michael C. Berch) writes: > >Informal estimates by people I know who are involved with the > >Internet Advisory Board and Internet Experimental Task Force put the > >number at more like 80,000-100,000. This is for *connected* sites. > > So this includes all the hosts on a local ethernet that can reach an Internet > gateway, then? I mean, Berkeley has something over 1,000 Sun workstations, and > those all can telnet or ftp to the Internet; so that makes them "Internet" > hosts? Yes. By definition. But not all hosts on LANs that can reach a system that is connected to an Internet network can telnet or ftp to the outside; the gateway may specifically be set up NOT to route packets between the Internet and the LAN, but would (for example) be set up to forward mail. This is the case at present at Sun and many other places. These hosts are not on the Internet for IP purposes, but are for mail purposes. This is the metric used to come up with the 200,000 number. > In which case, using the UUCP maps as a count of hosts is probably off by an > order of magnitude. AMD mentioned 600 nodes hiding behind their lone UUCP map > entry. Pyramid has over 250. There are scads of companies with fully connected > workstations and timesharing systems whose only tie to the rest of the world > is their one UUCP host. > > And of course those networks all use RFC-822 and talk SMTP. Most of them do. And that's why they're counted in the Internet group and not in the UUCP group. The purpose of my article was to compare the size of the RFC-821/822 community vs. the size of the V7 UUCP mail community. A lower bound on the size of the former is the connected Internet; an upper bound is the total number of systems speaking TCP/IP that have been sold and are still operating. I have trouble sizing the UUCP mail community, because so many sites in the UUCP Map are, as you point out, networks hiding behind a single published gateway. But, with the exception of AT&T, DEC, IBM, and a few others, a large number of those sites are really internal Internets that live in the RFC-821/822 world. Michael C. Berch mcb@ncis.llnl.gov / uunet!ncis.llnl.gov!mcb