ron%brl-vgr@sri-unix.UUCP (02/10/84)
From: Ron Natalie <ron@brl-vgr> We fixed it voluntarily, but... No where in the RFC810 (even the updated one) does it say what is legal with respect to CPU type or Operating system. Obviously you assume that these entries follow that of the "<name>". Not so, only the hostnames, netname, and gatename fields depend on this. It is clear that this isn't so. The examples for cputype list PDP-11/70 and C/30, neither of which conforms to the "<letter>[*[<letter-or-digit-or-hyphen>]<letter or digit>]" style. -Ron
rlb@RIACS.ARPA (02/14/84)
From: Bob Brown <rlb@RIACS.ARPA> The problem is simply that RFC 810 does not make explicit what are and are not valid for cpu-type and opsys. The "lexical grammar" section only gives examples and they in no way imply that "<name>" is the correct structure. The two things to be concerned about are 1) what syntax to use in forming entries to send to the NIC, and 2) what syntax to handle properly in receiving host table entries back from the NIC. These need not be identical. Notice that the BSD htable accepts slashes and dots in host names. This is fine. We shouldn't run syntax checkers on the NIC table just to have them break as they did a couple weeks ago. The rule-of-thumb to use is "be conservative in category 1" and "be liberal in category 2". In BRL's 68000 case, the error is in the second category above, hence Berkeley's problem - fix htable to be more robust. Bob Brown RIACS/NASA Ames ----------
feldman@tymix.UUCP (Steve Feldman) (02/20/84)
I'm the guilty one. I wrote htable a few summers back when I was a grad student at Berkeley working for the CSRG. It started out as something to do to play with YACC, and ended up being included in the distribution. I ran into these problems of inadequately specified syntax for some fields, and made sume assumptions to get the thing to work with the existing HOSTS.TXT. (That's why slashes work in cpu types.) I freely admit that the assumptions I made weren't general enough, and htable should be fixed to reflect reality. (Also, RFC810 should probably be fixed.) But I'm no longer at Berkeley, so I hope someone there is listening! By the way, htable actually did find some real syntax bugs in the host table, which the NIC did fix, although slowly. Steve Feldman ex-Berkeleyite
ron%brl-vgr@sri-unix.UUCP (02/22/84)
From: Ron Natalie <ron@brl-vgr> Good god! Why would anyone use YACC to parse HOSTS.TXT. It only requires a simple state machine. Try looking at anything on UNIX that parses /etc/passwd. The only difference is that HOSTS.TXT has spaces around the colons. -Ron