home *** CD-ROM | disk | FTP | other *** search
- From: gwyn@BRL.ARPA (VLD/VMB)
- Date: Sat, 29 Mar 86 0:57:42 EST
-
- I don't think P1003 should be trying to invent new facilities.
-
- [ I don't see anywhere in the original posting where such was suggested.
- Nor does an article appearing in this newsgroup imply that P1003 is
- going to act on it. See Volume 6, Number 2. -mod ]
-
- Suggestions like this should probably be forwarded to the
- UNIX-WIZARDS (net.unix-wizards) mailing list.
-
- [ Maybe. I was so pleased to see *anything* that wasn't on timezones....
- -mod ]
-
- In the case of writeeof(), suppose you write an EOF 1M bytes
- into a regular disk file, then write another 2M bytes into the
- same file, then read the file. According to the suggestion,
- the first EOF would be gone (implementation would probably
- also require this). But in the pipe example, multiple EOFs
- would stay around and be detected upon reading the pipe.
- (At least using stream i/o, one could conceivably slip an EOF
- control packet into the pipe stream, so this is implementable.)
- But look: Files and pipes would not behave the same; why
- weaken one of UNIX's strongest design strengths unnecessarily?
-
- I suppose one could argue more reasonably for the inclusion
- of a truncate() primitive, to shorten an existing file of
- any type; Fortran could use this. On the other hand, why
- encourage the use of Fortran on UNIX?
-
- Volume-Number: Volume 6, Number 8
-
-