brownell@harvard.ARPA (Dave Brownell) (08/29/84)
Insane Anglo Warlord <--> You Know Who!!
Everyone seems to blame USG for "breaking" UUCP ... 'cept it seems to
my fading memory as if the first UUCP I ever looked at (v7 before UCB)
had six character names. I don't have the sources here, so I can't
check. Is my memory lying to me about this, or was USG just being
consistent with the original?
On a more serious note, if anyone running the Sys V UUCP really
worries, it's easy to make it send out long names. I did it for ours.
The fixes were in uucpname.c, to fill up the buffer instead of
truncating, and in cico.c, to print the whole name instead of just the
first six characters. Two changes to uucpname.c, and one to cico.c, as
I recall. (Sorry, no diffs.)
Dave Brownell
Sequoia Systems Inc.
sequoia!brownell@harvard.ARPA
{allegra floyd ihnp4 seismo}!harvard!sequoia!brownellhoney@down.FUN (code 101) (08/31/84)
There are two aspects to the truncated sitename problem; your characterization of the lesser of them is on the mark. However, the other is by far the more serious of the two. Suppose my site talks to sdchema and sdchemb. Work files spooled for these sites are of the form C.sdchemaX1234 and C.sdchembY5678. I.e., the work file is of the form C. + sitename + grade + sequence where grade is one byte and sequence is four bytes. This leaves seven bytes for the sitename; anything longer is truncated. The horrible mistake made in System V UUCP was the decision to use two grade characters*, leaving but six digits for the host name prefix. This forces the above work files to be C.sdchemXA1234 and C.sdchemYA5678. Consequently, it is impossible for uucico to distinguish between work files spooled for sdchema and sdchemb. There seems to be no satisfactory resolution of this problem; while it is likely that most System V sites will ultimately run honey danber (which, as you may have guessed, reverts to the seven letter standard), there will *always* be a number of sites that don't upgrade. And this number may be significant, since lots and lots of companies have ported System V to obscure machines and don't feel pressured to upgrade. Peter * The appearance of the second grade character was motivated by a perceived need to reduce the number of times sequence numbers are retrieved (and the concomitant lock contention). Other versions of UUCP, notably 4.2BSD and honey danber, achieve the same goal in clever ways**. ** The mechanics can be found in gename.c (pronounced ge'-nam-ee-dot-c). It's interesting to note that 4.2BSD UUCP and honey danber use radically different techniques to effect the same goal. Even so, they were written by the same person***. *** Heh heh.
honey@down.FUN (code 101) (08/31/84)
As an aside, Seventh Edition UUCP used seven bytes in uucpname.c and cico.c. Peter
jwp@sdchema.UUCP (09/05/84)
In article <25@harvard.ARPA> brownell@harvard.ARPA (Dave Brownell) writes: >... if anyone running the Sys V UUCP really worries, it's easy to make it > send out long names... Unfortunately, Dave, this fails against the same sort of reasoning that is going to impose short identifiers on us. Specifically, we all have to be reduced to the least common denominator... To be explicit, Bell, or AT&T, or whatever name it's going under these days, has decided to force crappy thinking on the world, and all of us will have to conform to it, even though we have code that can handle this. Or source code so we could do what they should have done in the first place. You see, they've made a lot of money peddling binary-only licenses... Interestingly, at least some of the people who are going to shove this down our throats (by refusing to "register" sites with non-conforming names) were at UCB way back when... Then it was "You mean *you don't have a VAX*???? You mean you've only got an *11/40*??? With *RK05s*????? Well, that's too bad, buddy... We've got all this fine software, but if that's all you've got... Well... that's tough, man..." Now, of course, it's "Oh... You wanna run the UNIX standard on a Kaypro II? Hey, babe, don't worry!! We'll just lower the standard so you can say you conform to it." Sic semper venditori. Caveat Danaos... John Pierce, Chemistry, UC San Diego sdcsvax!sdchema!jwp