chuqui@nsc.UUCP (Chuq Von Rospach) (07/28/85)
Note: I've set this article up to move future discussions into net.mail, which is a more appropriate place for it. In article <433@olivee.UUCP> gnome@olivee.UUCP (Gary Traveis) writes: >Is anyone thinking of adding an override to the auto-routing >software running at a number of net-mail sites? I've looked at this as an offshoot of my playing with sendmail and routing here at nsc. There are two ways I can see of doing it within the normal mail naming space: o a pseudo-site. If the mailer sees a special site name, it knows to turn off path routing. Example: "nsc!no-opt!hplabs!foo!bar" doesn't get optimized. The problem is that we need a name that is a legal site name that doesn't turn into a real site name at some point in the future.... o a pseudo-domain. If the mailer sees a special domain name, it turns off path routing. Example: "nsc.NOOPT!hplabs!foo!bar" or "nsc!hplabs.NOOPT!foo!bar" depending on where you want to put it. Problems: lots, actually, including mucking around with the domain space in undefined ways and multiple domains on a single site (nsc.UUCP.NOOPT or nsc.NOOPT.UUCP?) Generally, I don't like either of these, because they muck with the naming space and paths in a non-standard way. What happens if someone who ignores the protocol routes it around someone who does? It could get really nasty. >It would make loop-back testing of net mail paths a whole lot >easier! I'm not at all sure that loop-backs are really neccessary, and the aims of a loop-back are directly against that of routing -- loop-backs require a specific path, and routing software is suppose to help hide a route from an address. If you really NEED a loop-back, you can get it through routing software in a couple of ways: o mail it to the SA at the site you're trying to loop to, and ask them to return the message to you. o mail to a non-existant site beyond the site you are trying to loop to, and let MAILER-DAEMON return it to you (as in ..!nsc!nothere!foobar) Neither of these require a way to get around the routing software, and I have never seen a reasonable reason beside loopback to be able to get around the routing software. Because of that, I've pretty much given up looking for a way to disable it in a way which works on the network as it currently stands. -- :From the carousel of the autumn carnival: Chuq Von Rospach {cbosgd,fortune,hplabs,ihnp4,seismo}!nsc!chuqui nsc!chuqui@decwrl.ARPA Your fifteen minutes are up. Please step aside!
peter@kitty.UUCP (Peter DaSilva) (07/31/85)
> Neither of these require a way to get around the routing software, and I > have never seen a reasonable reason beside loopback to be able to get > around the routing software. Because of that, I've pretty much given up > looking for a way to disable it in a way which works on the network as it > currently stands. Here's a good reason: with the network growing at the rate it is, and with all sorts of other changes occuring in the net (new links showing up, old links disappearing, etc...), nobody can keep all their routing info completely up to date. One of the main feeds for both of the sites I'm the contact for frequently screws up the routing. In addition, the routes it chooses are occasionally non-reflexive, especially with the old V7 mailer I'm reduced to using... Luckily I can get around the routing by giving it an explicit route. It only calculates the route for the first hop, you see. If it optimised everything I'd be completely lost. Like it or not, the net is a dynamic structure. Like it or not, there are sites that have old software. Unless you're going to distribute a PD mailer that will run on a vanilla PDP-11, I'd strongly discourage the widespread use of route optimisations.
sob@neuro1.UUCP (Stan Barber) (08/03/85)
In article <184@kitty.UUCP> peter@kitty.UUCP (Peter DaSilva) writes: >One of the main feeds for both of the sites I'm the contact for >frequently screws up the routing. In addition, the routes it chooses are >occasionally non-reflexive, especially with the old V7 mailer I'm reduced >to using... > "One of the main feeds" he refers to is neuro1, the site I administrate. I would note the we were getting a large number of rejection to messages Peter sent though this site because his routes were 10-15 site long and were frequently truncated and then rejected. As to "screw"ing up the routing, I am using the material provided by the UUCP project for any routing information. If you want to insure that routing information is correct, send your map information to uucpmap@cbosgd.ATT.UUCP. >Luckily I can get around the routing by giving it an explicit route. It only >calculates the route for the first hop, you see. If it optimised everything >I'd be completely lost. Nothing lucky about it. It was planned that way so that people who like to use explicit route can. The whole point (in my opinion) behind an intelligent mail system is to allow unsophisticated users to be able to send mail with minimal information (username and host and as time goes by, domain), while allowing sophisticated users to use explicit paths if they so desire. >Like it or not, the net is a dynamic structure. Like it or not, there are sites >that have old software. Unless you're going to distribute a PD mailer that will >run on a vanilla PDP-11, I'd strongly discourage the widespread use of route >optimisations. I disagree with this. I think as all networks become more complex, route optmization will be the ONLY way for a casual user to be able to use the networks. Since there will be more and more casual users, we ( the people that worry about mail, routing and so on) must be able to accomodate them to make it as easy as possible to them to use mail, news and the like. Peter, if you don't like mailing through neuro1, I can bounce all your mail back...:-) -- Stan uucp:{ihnp4!shell,rice}!neuro1!sob Opinions expressed Olan ARPA:sob@rice.arpa here are ONLY mine & Barber CIS:71565,623 BBS:(713)660-9262 noone else's.