matt@oddjob.UChicago.UUCP (Matt Crawford) (02/17/85)
A summary for those who do not wish to read the ~60 lines that follow: This is a discussion (bordering on a flame session) about some details of the proposed standards which have come out of the "UUCP Project". See also article <597@plus5.UUCP> for another criticism. Certainly it is a Good Thing that people are taking the trouble to organize the anarchy that is uucp (while leaving untouched the anarchy that is usenet!) it will be a Big Disappointment if the result of the project contains enough misfeatures that it is not widely adopted. In article <ulysses.1109> smb@ulysses.UUCP (Steven Bellovin) writes: >> From: matt@oddjob.UChicago.UUCP (Matt Crawford) >> Haven't you heard? Mark Horton, on behalf of the "UUCP Project", >> wants us all to keep our host names down to 6 characters. Why? I >> don't really know, but I understand that a certain large unix vendor >> sold a binary-only uucp implementation which only allows six char- >> acters. This could be coincidence of course. > >Do you think you could possibly control your paranoia for a while? The >six-character limit (which, by the way, is that the name be unique in >the first six characters, not that it be six characters only) came with >System V, about two years ago. That far antedates the UUCP project. That is my point exactly. The UUCP project seems to be catering to AT&T's system V implementation of uucp. In addition, the "UUCP Mail Transmission Format Standard" (hereafter "MTSF") by Mark Horton uses the phrase "6 letter UUCP [site] name", never saying that names might be longer if they are unique to 6 characters. Looking even further, The document "UUCP Subdomain Requirements" by Mark Horton and Karen Summers-Horton suggests, when choosing site names: "Plan for the day you have lots of machines, for example, ``framus-a'' or ``a.framus'' if your company name is Framus and your theme is letters of the alphabet, ..." Either the framus people have to break the rule or have to create a subdomain of one machine to allow for expansion later. >What the hell do you want Mark to do (and it's not just Mark, incidentally; >there was a large group involved in discussing the subject), pick a >solution that will break half the machines on the net? The Prime Directive >of the project was that no solution would be considered that demanded >universal co-operation, because we know damned well that we *can't* (not >won't, can't) get it. To quote the MTFS again: "(Note - if an intermediate site adds text such as ``system!'' to the front of a ``user@domain'' syntax address, ... this is a violation of the standard.)" No need to worry about breaking half the machines on the net -- 90% are already broken! It seems to me that the UUCP project will require some new software to be sent around for free. I also thought this was in fact intended to be done. Why not send out a uucp that is more flexible as part of the package? >The six-character truncation was a mistake. Everyone involved agrees. >So what did AT&T Bell Laboratories do? We fixed it. The new uucp (called >honey danber, for its authors) is what's being run on ihnp4, ulysses, cbosgd, >and a host of other sites. I'm told it's available to the outside world >now, through the usual AT&T Technologies licensing channels. For $5000 (source) according to a recent net article. _____________________________________________________ Matt University crawford@anl-mcs.arpa Crawford of Chicago ihnp4!oddjob!matt
mark@cbosgd.UUCP (Mark Horton) (02/18/85)
In article <606@oddjob.UChicago.UUCP> matt@oddjob.UUCP (Matt Crawford) writes: >That is my point exactly. The UUCP project seems to be catering to >AT&T's system V implementation of uucp. If you have an alternative that will work, we'd love to hear about it. After extensive discussions with a number of UUCP experts, we concluded that we have no choice but to require all transport layer names to be unique in the first 6 characters. Among the reasons for this are that all System V release 1 and 2 systems in the world have this 6 character truncation, that System V Release 3 will not be released for some time, and that even after all the major suppliers (e.g. Berkeley and AT&T, the two roots from which the other vendors base their distribution) are distributing a 14 character version, it will be years and years until the rest of the world has converted. There are still machines out there running V7, and System III is still a major part of the market. If one machine on the network only looks at the first 6 characters, they can't talk to two machines that are not unique in the first 6, which in turn means that these two machines can't safely choose to use the same first 6 letters in case they might have to talk to a machine that has this problem. I don't think we can assume 14 characters until about 1990. >In addition, the "UUCP Mail >Transmission Format Standard" (hereafter "MTSF") by Mark Horton uses >the phrase "6 letter UUCP [site] name", never saying that names might >be longer if they are unique to 6 characters. This is just naming. We had to call these things something. Perhaps we should call them "transport addresses" or something. Looking even further, >The document "UUCP Subdomain Requirements" by Mark Horton and Karen >Summers-Horton suggests, when choosing site names: > > "Plan for the day you have lots of machines, for example, > ``framus-a'' or ``a.framus'' if your company name is Framus > and your theme is letters of the alphabet, ..." > >Either the framus people have to break the rule or have to create a >subdomain of one machine to allow for expansion later. This is in reference to the domain names, not the transport names. Domain names are not restricted to be unique in the first 6 chars. >Why not send out a uucp that is more flexible as part of the package? Mainly because UUCP is proprietary to AT&T and we'd get sued if we did. We also have to deal with machines that have binary licenses and so we can't send them modifications to existing code. Finally, there are restrictions in the System V license that say that even if the recipient has a System V license, we have to pay AT&T some huge sum to become a distributor and then pay AT&T a $750 royalty on every copy of our stuff we send out, as though it were a full UNIX, if it contains any part of System V. Being a not-for-profit project with no working funds and a volunteer structure, this is not possible. Mark Horton
henry@utzoo.UUCP (Henry Spencer) (02/19/85)
C'mon guys, this is silly. What Dave was saying was "please maximize the
information content, not the cuteness, of your site names". This is a
reasonable principle *regardless* of how one feels about domain addressing
and other holy wars. Unless it's strictly for internal consumption and
the software is arranged so that it will never be seen by the net at all,
both "snow" and "white" are very poor choices for machine names.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry