Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site pur-phy.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxj!houxm!ihnp4!inuxc!pur-ee!CS-Mordred!Pucc-H:pur-phy!crl
From: crl@pur-phy.UUCP (Charles LaBrec)
Newsgroups: net.news.b
Subject: Re: More strangeness from 2.10.2
Message-ID: <1484@pur-phy.UUCP>
Date: Tue, 23-Oct-84 12:18:08 EDT
Article-I.D.: pur-phy.1484
Posted: Tue Oct 23 12:18:08 1984
Date-Received: Thu, 25-Oct-84 02:46:07 EDT
References: <1464@pur-phy.UUCP> <11@uwvax.UUCP>
Reply-To: crl@pur-phy.UUCP (Charles LaBrec)
Distribution: net
Organization: Purdue Univ. Physics Dept., IN
Lines: 48
Summary: 


I have finally taken some time to look at 2.10.2, and have decided that
the way sortactive() works is undesirable, to say the least.  I really
dislike new groups being put at the top.  I set up my .newsrc to see
IMPORTANT groups there, and just because some unimportant one that I
want to read comes in, I have to edit the .newsrc.  What a pain.  Furthermore,
it sorts them in a semi-random order sometimes.  Here is a good example.
A new user without a .newsrc starts reading, does a few groups, and quits.
The next time he/she starts, the next batch will be placed at the beginning,
circumventing the active file order, which at least is reasonable.  By the
end of a few sessions, his/her .newsrc will be a mess.  If new groups were
added to the end, the default would be to keep the .newsrc in active file
order, more or less.  What I meant about semi-random is this:  the function
that does the compares for qsort() in sortactive() only compares the line
numbers that that group is found in the .newsrc.  When a group is not in
the .newsrc, its line number is considered to be -1.  So all groups not
listed compare equally.  Consider the following .newsrc and active files:
.newsrc:			active:
	general:			general
	e:				a
	b:				b
					c
					d
					e

After the sort, I think the active would look like:
	general
	e
	b
	c
	d
	a
So what was once 2nd is now last.  At our site, this means a local group
would get put at the end.  With a longer .newsrc, it would really jumble it.

Since the compare routine gets called with a struct pointer that contains
both the newsrc line index and the active file position, I intend to check
for an index of -1, and then sort by active file pos.  Does anyone out
there see a problem with this?

About the man pages, for one I notice that checknews has a new option -N
to check only specific groups.  I'd really appreciate a new set of all the
man pages so that I don't (and my users don't) have to look at the source
to determine what is new.  (Expire also has a new switch or two.)

Charles LaBrec
UUCP:		pur-ee!Physics:crl, purdue!Physics:crl
INTERNET:	crl @ pur-phy.UUCP