home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.protocols.tcp-ip.domains
- Path: sparky!uunet!elroy.jpl.nasa.gov!ames!dscs.arc.nasa.gov!medin
- From: medin@dscs.arc.nasa.gov (Milo S. Medin)
- Subject: Re: Why no MX records for .BITNET hosts?
- Message-ID: <1992Dec28.045444.3467@news.arc.nasa.gov>
- Sender: usenet@news.arc.nasa.gov
- Organization: NASA Science Internet Office
- References: <VIXIE.92Dec26031209@cognition.pa.dec.com> <DWELLS.92Dec27141205@fits.cv.nrao.edu> <VIXIE.92Dec27122627@cognition.pa.dec.com>
- Date: Mon, 28 Dec 1992 04:54:44 GMT
- Lines: 47
-
- Well, we run the dnet.nasa.gov MX hack and associated forwarder. We also
- have a similar hack for span.nasa.gov that also works. I agree that it's
- a very convienent hack, but also agree that architecturally it's the wrong
- thing. Many if not most of the DECNET systems on the NSI-DECNET (what
- SPAN was absorbed into) have IP connectivity as well as DECNET. They
- maintain DECNET access to talk to DECNET only sites and support
- applications that depend on DECNET transport.
-
- The reason *.dnet.nasa.gov is convienent is obvious; the reason it's
- a kludge is because mail takes a suboptimal path for delivery to systems
- which are dual homed (IP and DECNET). The mail relays for these fake
- domains pull a fair amount of load. And the DECNET users have to manually
- route mail through the relays because DECNET mail doesn't support things
- like MX records (of course, it also doesn't support multiple names per
- node address or mail spooling as shipped either, so users aren't expecting
- much I guess).
-
- The right thing to do is to install individual MX records for the DECNET
- nodes that have native IP support, and maybe for all nodes at a site
- through a local mail relay. This would reduce the load on the "default"
- mail relays, and provide a more direct mail path as well. This was the
- initial plan in fact when we turned this service on, but it's a pain
- to support since the MX records have to be manually configured, and
- the poor DECNET only users now have to know what locally specific syntax they
- have to use for Internet relay. This makes it difficult to write common
- user services documentation, etc...
-
- Anyways, it's proven so popular that it's just easier to try and harden
- the mail relays so this works, and inform the users about the limits of
- this approach. It also winds up using more DECNET bandwidth than
- needed, but these days most if the DECNET rides over the same trunks
- that the IP traffic does via multiprotocol routers, so I guess it's
- not that big a deal anyway.
-
- The same may be true re: BITNET actually, since I gather most of the large
- BITNET sites get their trunking via the Internet anyway, using this
- RSCS over IP stuff. I would expect the amount of BITNET traffic overall
- to be dropping off, similar to what's been happening to DECNET, as
- more hosts convert to native IP support (or the machines themselves are
- scrapped and sent to surplus - anyone want a few 780's cheap? :-))..
-
- Thanks,
- Milo
-
- PS The above of course is not the official position of anyone, much
- less the US Government, etc, etc...
-
-