tp@ndm20 (08/10/85)
Please don't flame me, I think I'm going to make a reasonable point
here.
Does anyone else on the net happen to have a mailer that doesn't
know what a header is? That only knows one form of routing
(<next-host>!<feed-this-to-next-host-and-let-him-figure-it-out>)?
Let me describe my situation by way of example.
I am on a Sys V machine. My only mailer is mail(1). I don't have
sendmail. Actually, I believe that is berkely only. I don't have
a source license. My only neighbor is a site that also happens to
be on CSNET. His mailer gives @ precedence over !. I have no control
over this. If domains are adopted, I will be cut off from mail until
such time as there is a UUCP domain server on either arpa or csnet,
because mail to mark@cbosgd.att.uucp will look to a csnet site like
something to foist upon arpanet for delivery.
Pathalias has the critical advantage of being compatible with existing
reality. I can generate a path and feed it to my neighbor. I don't
happen to have pathalias (everyone tells me "Oh, it will be posted again,
just wait for it"), and I don't have a mailer that can be made to use it
to any great degree (yeah, I can write a front end shell script, but I'm
talking about a little better functionality). Yet I can still send mail,
receive notes (news to the rest of you), and generally exist on the net.
If domains come into existance, it will fragment the net into pieces that
can not communicate, until such time as all systems adopt the new software
(which of course must first be ported to all existing systems, and would
of course then be distributed free by all vendors, so that a new node could
join the net).
Those who advocate domains, tell us how you are going to accomplish the
transition. Neither you nor I can force my neighbor to change his software
if he doesn't want to, so how do I stay connected to the net (if I upgrade;
obviously since my mail(1) program doesn't do headers, I have to upgrade to
stay on the net). Or is the tactic to design the new software so that the
only way to stay on the net is to upgrade?
Then we will have a domained usenet, and a usenet using existing software
that will reconfigure itself as hosts switch to the new software (neat
how usenet is flexible enough to do that, isn't it?). Then somebody will
build gateway software, because he wants to talk to both kinds of hosts.
Now we have a new domained network, and good old usenet, smaller, but of
course it will grow because new unix machines already have the software to
run it, and it still has all the same problems you thought you just solved.
The point I want to make is that we are not designing a new network, we are
talking about retrofitting an existing one. Sure arpa did it, but they could
hire people to write software and coerce people to use it. Usenet can't,
because many of us can't pay a $50,000 membership fee to support such efforts.
(Even $15,000 for half of a CSNET membership.)
Whatever is done must be implementable. I don't see domains as being such.
(Look at the summaries that get posted of which sites are running which version
of the notes and news software if you think people will adopt it because it
is 'better'). It must be compatible with existing systems and software.
Good Luck
Terry Poot
Nathan D. Maier Consulting Engineers
(214)739-4741
Usenet: ...!{allegra|ihnp4}!convex!smu!ndm20!tp
CSNET: ndm20!tp@smu
ARPA: ndm20!tp%smu@csnet-relay.ARPAjsq@im4u.UUCP (08/16/85)
In article <3700002@ndm20> tp@ndm20 writes: ... >I am on a Sys V machine. My only mailer is mail(1). I don't have >sendmail. Actually, I believe that is berkely only. I don't have >a source license. My only neighbor is a site that also happens to >be on CSNET. His mailer gives @ precedence over !. I have no control >over this. If domains are adopted, I will be cut off from mail until >such time as there is a UUCP domain server on either arpa or csnet, >because mail to mark@cbosgd.att.uucp will look to a csnet site like >something to foist upon arpanet for delivery. ... You will probably find that your System V /bin/mail does know about the UNIX From space line, if it has a reply command. It doesn't know about RFC822 headers, but it doesn't have to. Mail to neighbor!ihnp4!cbosgd.att.uucp!mark, where neighbor is your neighboring site which is also on CSNET. It's a pure UUCP address syntactically (all bangs, no ats). There's no reason the CSNET site should interpret it as anything else. (Also, uucp is not a valid top level domain in the Internet, so there's no reason a CSNET host should send mail for cbosgd.att.uucp to the ARPANET anyway.) There is no compatibility problem, and no fragmentation of the UUCP network. Domains and old-style routing co-exist. You don't need any new software. Your CSNET neighbor doesn't need any new software. Many people will, however, get new software, because they will find domains work better than source routing. -- John Quarterman, UUCP: {ihnp4,seismo,harvard,gatech}!ut-sally!jsq ARPA Internet and CSNET: jsq@ut-sally.ARPA, soon to be jsq@sally.UTEXAS.EDU
peter@baylor.UUCP (Peter da Silva) (08/16/85)
> Please don't flame me, I think I'm going to make a reasonable point > here. > > Does anyone else on the net happen to have a mailer that doesn't > know what a header is? That only knows one form of routing > (<next-host>!<feed-this-to-next-host-and-let-him-figure-it-out>)? Yo! Here I am! Most commercially available systems are in this boat. Why not just let the gateways do whatever magic they need to get '!' formats through their nets. Convert it to @sys,@sys,@sys:foo@sys if you need to, but change it back when it gets to usenet. I just got a letter from someone on some random net in the uk, and I have no earthly idea what priority the systems in the way give to '!' and '@' and '%' and '.'... I don't think there is a way to know. Then us plebes can maybe get some mail back to people who ask us questions. -- Peter da Silva (the mad Australian werewolf) UUCP: ...!shell!neuro1!{hyd-ptd,baylor,datafac}!peter MCI: PDASILVA; CIS: 70216,1076