Re: [bug-inetutils] some notes on inetutils-1.8

[melodramus]

hello alfred 

somehow you don't understand me well. i'll try to be more explicit.

On Tue, 17 Aug 2010 01:58:02 -0400
"Alfred M. Szmidt" <address@hidden> wrote:

>    however, the configuration was painful. i tried the easy way,
>    turned off servers and clients and wrote --disable-syslogd=no
>    etc. but that wasn't supported. thus i needed four lines to get the
>    right selection.
> 
> The following should work to compile only syslogd, as per ./configure
> --help, INSTALL and README.
> 
> ./configure --disable-server --disable-clients --enable-syslogd
> 
> What `four lines' did you use?  Maybe something can be improved.

i just disabled everything i didn't want, which made up four text lines
(rows?) on my monitor. if --enable-syslogd works, --disable-syslogd=no
should work too (by convention), like --disable-syslogd should be
substitutable with --enable-syslogd=no. using --disable-syslogd=no has
the advantage that one can middle-click (Xorg quick copy/paste) the
option from the configure --help output.

> 
>    further, why is encryption and authentication chained to kerberos?
>    does this have to be this way? i'm not in that.
> 
> Do you have a different scheme in mind, what if so?  We like kerberos,
> since it is a technically good solution.

i'm not in that but kerberos is from m*soft and not that typical on
other machines. on linux nearly all free software uses openssl or
gnutls. i don't have kerberos on my system and don't want it,
especially not for one single small package.

>    in the end, installation went fine (though i didn't test yet)
>    except of syslogd not being installed to the target directory. i
>    moved it manually.
> 
> What do you mean?

what i wrote! after make install, syslogd did not reside at the target
location. i copied (or moved (??); have forgotten.) it manually
from src/ to /bin.

> 
>    i also miss a man page for syslog.conf. because you added extra
>    symbols, there should be an own man page. not to say that there are
>    so many versions of syslogd out on the net that the man page should
>    mention the origin in the first paragraph. on the other hand, i
>    believe that the info document is unique. however, ...
> 
> It isn't unique in the world of GNU, it is our prefered format for
> documenting our programs, we can include a chapter in the man page for
> syslogd if that is of help.  Would you like to submit such a patch?
> But the main documentation will always be the info manual.

you got me wrong. i thought that no other syslogd package provides an
info document so that the reference to the origin isn't that important
than in the man page. however, i found that sysklogd, which i used
until today, installed info documents as well. because inetutils does
not come with an info document for syslog.conf, i now have one info
document for inetutils syslogd and one for sysklogd syslog.conf on my
system. you can imagine that both don't fit in all details.

the point is that on *nix it is common to document all entities
separately. try 'man hosts' or 'man resolv.conf' (you can even try 'man
utf-8' if you don't know about this format.) please stay in line with
this tradition and don't treat man pages like toilet paper. there are
users who think that the info system is rubbish and like the charm of
good old man pages. also, please refer to the origin (inetutils) in the
first paragraph because there are too many incompatible versions out
there, and the admin definitely wants to know if the man page he reads
is really the one reflecting the tool he uses (some distros ship
standard manuals, which are found instead of the ones freshly
installed. also, one may have forgotten to clean up old manuals.)

> 
>    possibly syslogd should better be moved to util-linux, because it
>    isn't really an inet util (that it can forward doesn't compare it
>    to tools like ftp and friends.)
> 
> Then non-Linux kernel using operating systems would not have a
> improved version of syslogd, like for example GNU/Hurd, Solaris,
> OpenBSD, et al.  syslogd is best fitted in inetutils, hence why we
> provide it.

ok, understand that. still find it displaced in inetutils and
would expect it as an own package, as it always was. but times
change, it seems...

>    hostname, instead, is already in the coreutils package.
> 
> The idea is to remove hostname from coreutils; if I recall it is not
> compiled per default in coreutils anymore.
> 
>    if inetutils is intended to make net-tools redundant, i'd like to
>    see route and others as well. but you have that on your list, i
>    believe.
> 
> The intention of inetutils to provide all relevant network tools for
> GNU systems and non-GNU systems.

so, route, netstat and friends are on the list!

> 
>    what about an iproute implementation (without berkeley-db
>    dependency ;)
> 
> Would you like to write such a program for us?

would i then ask you for doing it? i'm not in this field of
programming, just maintain a system. sorry.

> 
>    (btw., choosing -? for help and -h for hop on the syslogd command
>    line is not a good choice in my eyes. please rethink!)
> 
> You will need to have a stronger argument than "I don't like it".

just direct your argument against yourself and ask yourself what
'strong argument' you had for breaking with the tradition/convention.
what you did is inconsistent with thousands of other command-line tools
for *nix (and even for w*ndows) and thus a pain in the ass for every
admin with weak memory!

best wishes,
MeloDramus <address@hidden>