Re: [bug-inetutils] ping, --count=NUMBER

[Sergey Poznyakoff]

Alfred M. Szmidt <address@hidden> ha escrit:

> Thanks for the explanation.  `--count 4' has never been valid with
> argp.

Not at all, it is perfectly valid:

address@hidden:~$ ping --count 4 gnu.org
PING gnu.org (199.232.41.10): 56 data bytes
64 bytes from 199.232.41.10: icmp_seq=0 ttl=47 time=176.773 ms
64 bytes from 199.232.41.10: icmp_seq=1 ttl=47 time=176.617 ms
64 bytes from 199.232.41.10: icmp_seq=2 ttl=47 time=176.241 ms
64 bytes from 199.232.41.10: icmp_seq=3 ttl=47 time=176.600 ms
--- gnu.org ping statistics ---
4 packets transmitted, 4 packets received, 0% packet loss
round-trip min/avg/max/stddev = 176.241/176.558/176.773/0.195 ms

With long options, both `--opt=arg' and `--opt arg' are valid constructs.

> As for the `-c 4' vs. `-c4' case, since we already use short-options
> with optional arguments, the behaviour is already weird.

Well, I'm not saying it's weird.  Getopt makes its best to make it
deterministic at least.  But the downside is that one must remember
which option takes mandatory and which one takes optional arguments,
to correctly use them.  Of course, we do use optional arguments when
it is reasonable and does not harm compatibility.  But in this
particular case I have strong doubts:

> In other words, users already must think about if they are using a
> short option with an optional argument, or not so they cannot depend on this
> behaviour.

Users must always think first, that's a good rule :^)  That's not a
problem.

The problem with `-c', as I see it, is that we are changing the behavior
of a traditional option that was in ping from the days of its inception and
many people had already made their habits of using it.  Often these
habits include separating the option argument by a space, because it
improves readability.  As a side note, at least one GNU manual encourages
users to use this style of passing option arguments: it is the GNU tar
manual.  I am quite sure, many people use ping in their scripts this
way, that is with `-c' and its argument separated.  The proposed change
will break those scripts.  That is the most important reason that speaks
against this change.

Instead of changing the `-c' behaviour we can simply introduce a new
option that will cause ping to stop after the default number of packets.
FWIW, the `-C' letter is still unused in ping, so we can use it for
that purpose.  This will make for a good mnemonics.  The corresponding
long option might be more verbose, say `--default-count'.

What do you think?

Regards,
Sergey