Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84 SMI; site sun.uucp
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!mtuxo!mtunh!mtung!mtunf!ariel!vax135!petsd!pesnta!amd!amdcad!decwrl!sun!guy
From: guy@sun.uucp (Guy Harris)
Newsgroups: net.micro.att
Subject: Re: ATT 3b2 (SVR2) questions
Message-ID: <2471@sun.uucp>
Date: Tue, 23-Jul-85 07:04:33 EDT
Article-I.D.: sun.2471
Posted: Tue Jul 23 07:04:33 1985
Date-Received: Thu, 25-Jul-85 08:05:28 EDT
References: <746@wlcrjs.UUCP>, <373@cuae2.UUCP> <513@bu-cs.UUCP>
Organization: Sun Microsystems, Inc.
Lines: 43

> "A read will not be satisfied until at least MIN characters have been
> received or the timeout value has expired...The MIN and TIME values are
> in the position for the EOF and EOL characters respectively. The time value
> represents tenths of seconds."
> 
> Not too clear is it: Does a zero value in TIME produce non-blocking I/O
> (timeouts occur immediately?) or should that be -1?  or does that end up
> to be 255 on most machines and then the read times out in 25.5 seconds?

The only real explanation of what (V)MIN and (V)TIME do that tells
*everything* is in the System V Interface Definition; it gets a subsection
to itself which is 5 paragraphs long.  I think the System III rules were
quite fine, thank you - they took less space to explain, and the only thing
they didn't do is provide a way to do read timeouts, but that's what "alarm"
is for.

Basically, if MIN is non-zero:

	If TIME is zero, there is no timeout of any sort; the read
	doesn't complete until (int)c_cc[VMIN] characters have come in.

	If time is non-zero, the read completes if (int)c_cc[VMIN]
	characters have come in or *at least one* character has come in
	and (int)c_cc[VTIME] 10ths of a second have elapsed since the
	first character came in.

If MIN is zero:

	If TIME is zero, you get a no-delay read (that's what the S5ID says,
	at least; I haven't checked the code).

	If TIME is non-zero, the read completes as soon as a character
	comes in or (int)c_cc[VMIN] 10ths of a second have elapsed since
	the "read" system call was first made.

> And I won't even mention how bad an idea it was to put termio into the
> administrator's manual (hasn't anyone at ATTIS heard of students?)

AMEN.  WAKE UP, SOMEBODY, AND PUT THAT SECTION IN THE *P*R*O*G*R*A*M*M*E*R*S*
MANUAL WHERE IT BELONGS!  Administrators don't care how to write screen
editors.  Programmers do.

	Guy Harris