Prev: Processing stdin in blocks
Next: Bash directory traversal with spaces
From: LC's No-Spam Newsreading account on 3 Mar 2010 05:17
On Tue, 2 Mar 2010, pk wrote:

> Usually it's a matter of enabling SNMP on the switch (usually it's enabled,

It's present but not configured. We (neither me nor the institute net
administrator) never bothered to learn that

> these days). In that case, based on what you said, I'd go with telnet first,
> since it's easily scriptable, and keep http as a last resort.

Hmmm ... telnet is scriptable at the price of using expect (which I
had to use only twice in my life, one being the other swith case). In
the case of the new switches I can't find the relevant info in the
telnet interface. Also the script is last resort, I am phasing out
the scripts for interactive use (I'd probably keep them for
crontabs) in favour of java, and prefer to do HMTL parsing in java

> However, consider that the html you get from the device [...]
> or it may contain javascript, or....

Yeah, I'm afraid my trouble could be due to imbedded javascript.

If I cannot overcome it, I'll explore snmp (in the worst case for java
I'll hook it to an external script, as I already do for arping)


--
----------------------------------------------------------------------
nospam(a)mi.iasf.cnr.it is a newsreading account used by more persons to
avoid unwanted spam. Any mail returning to this address will be rejected.
Users can disclose their e-mail address in the article if they wish so.
From: pk on 3 Mar 2010 06:00
LC's No-Spam Newsreading account wrote:

> crontabs) in favour of java, and prefer to do HMTL parsing in java

That makes it slightly better, as using a parser is (or should be) a more
robust option for parsing html.

>> However, consider that the html you get from the device [...]
>> or it may contain javascript, or....
>
> Yeah, I'm afraid my trouble could be due to imbedded javascript.

I was speaking out of personal experience. I once had a device (years ago)
that used a peculiar authentication mechanism. Essentially, the
authentication page contained a nonce buried into some javascript code. The
javascript code, in turn, was associated to the "submit credentials" button.
The user entered username and password, and upon submission the javascript
code would somehow encrypt the username+password pair (or just the password,
I don't remember now) using the nonce, and the result was POSTed to the
device.
So the POSTed data was different every time.

They had been using a script for a long time to automate logins to the
device and get to a diagnostic page. This script was using a complicated
patchwork of commands, using wget to download the authentication page, sed
or grep (or possibly both) to extract the nonce from the javascript, some
Perl to replicate the encryption algorithm (that somebody had translated to
Perl after reading the javascript code) and compute the data to POST, and
finally wget again to send the actual authentication data and to request the
diagnostic page. Then again Perl to parse the HTML and extract the actual
information the script was supposed to gather. If I'm not mistaken, the
device was also setting a cookie at some point, which of course the script
had to manage.
I don't remember the details, but the script, even if it mostly worked, was
quite long, clunky and required some time just to understand what it was
doing.

With SNMP, that was replaced by a 10-lines script using snmpwalk and snmpget
to get the same information and Perl to parse it. And that was due to my
lazyness, as doing it all in Perl only, using a SNMP module, would probably
have been even shorter.

> If I cannot overcome it, I'll explore snmp (in the worst case for java
> I'll hook it to an external script, as I already do for arping)

See if the above helps. Look at the authentication form HTML source, and see
if there's something weird in there. Sniff the data that goes to/from the
switch, and see if there's something that you missed or did not foresee in
your script.

(I still maintain that SNMP is cleaner and easier, but I see your point
about having to move to something new)
From: mop2 on 3 Mar 2010 13:29
On Wed, 03 Mar 2010 07:17:18 -0300, LC's No-Spam Newsreading account
<nospam(a)mi.iasf.cnr.it> wrote:
> Hmmm ... telnet is scriptable at the price of using expect
> (which I
> had to use only twice in my life, one being the other swith
> case). In
> the case of the new switches I can't find the relevant info in
> the
> telnet interface. Also the script is last resort, I am phasing
> out
> the scripts for interactive use (I'd probably keep them for
> crontabs) in favour of java, and prefer to do HMTL parsing in
> java
>


You can try html with shell:
x=0;while read -d\>;do printf "<$x>${REPLY%<*}";x=$[x+1];done



My option for telnet in the past without expect, bash only:


#!/bin/bash
exec 2>>/tmp/autoIP.log
user='' # user's login name (modem)
pass='xxx123yyy' # user's password (modem)
ppp=Tel # ppp connection name (modem Alcatel Speed Touch Pro)
#Log=/tmp/autoIP.log

d="" ### Debug authentication? "1" = yes "" = no

exec 3<>/dev/tcp/10.0.0.138/23

Auth(){ read -d$1 -r L<&3;[ $d ]&&echo -n "$L : ">&2; echo -ne
"$2\r" >&3;}

Send(){ x=0
while true;do b=${1:$x:1}; [ "$b" ]||break
echo -n "$b" >&3; read -d"$b" r<&3; x=$[x+1]
done
echo -ne "\r" >&3; read -d"$2" r <&3; [ "$3" ]&&echo "$r"
}

getIP(){ Read=''
while read N; do
[ "$Read" ]&&N=${N#*:}&&echo -n ${N%% *}&&break
[ "${N:6:16}" == "$1" ]&&Read=1
done
}

Auth ":" "$user"; Auth ":" "$pass"
read -t 1 -d\> -r L<&3;[ $d ]&&echo -n "$L>">&2



st=0;ppp="$ppp "; ppp=${ppp:0:16}

case $1 in
IP)Send "ip aplist" ">" o|getIP "$ppp"
;;nat)
Send "$1 $2" "="
Send "tcp" "="
Send "$3" "="
Send "$4" "="
Send "$5" "="
Send "$6" "=" o|\
while read L;do [ "${L:1:6}" == "Failed" ]&&st=1;done
#[ "$2" == "delete" ]&&Send "dns delete index=0"
[ "$2" == "create" ]&&{
#Send "dns add hostname=$7 addr=$5" ">" # endereço externo
#Send "dns add hostname=$7 addr=$3" ">" # endereço interno
#Send "dns domain domain=$8" ">"
#=======
# NO ANTIGO autoIP:
# [ "$IPa" ] && /sh/telnetAlcatel nat delete 10.0.0.2 1080 $IPa
80
# /sh/telnetAlcatel nat create 10.0.0.2 1080 $IP 80 $Srv $Dom ||\
# com template 0.0.0.0 nao é necessario, o proprio modem cuida disso
# /sh/telnetAlcatel nat create 10.0.0.2 1080 0 80 $Srv $Dom ||\
# echo "Houve erro na montagem da tabela NAT no modem" >&2
#========
echo -ne "\n`date +%y%m%d%H%M%S` telnetAlcatel:\n" >&2
Send "nat list" ">" o |grep $3|grep t|\
/usr/bin/tr -d [:cntrl:]|/usr/bin/tr -s " " >&2
echo >&2
}
esac

exec 3<&- # close connection

exit $st
From: Janis Papanagnou on 3 Mar 2010 17:46
mop2 wrote:
> On Wed, 03 Mar 2010 07:17:18 -0300, LC's No-Spam Newsreading account
> <nospam(a)mi.iasf.cnr.it> wrote:
>> Hmmm ... telnet is scriptable at the price of using expect (which I
>> had to use only twice in my life, one being the other swith case). In
>> the case of the new switches I can't find the relevant info in the
>> telnet interface. Also the script is last resort, I am phasing out
>> the scripts for interactive use (I'd probably keep them for
>> crontabs) in favour of java, and prefer to do HMTL parsing in java
>>
>
>
> You can try html with shell:

ITYM "bash shell".

> x=0;while read -d\>;do printf "<$x>${REPLY%<*}";x=$[x+1];done
>
>
> [...]
From: mop2 on 3 Mar 2010 21:31
On Wed, 03 Mar 2010 19:46:56 -0300, Janis Papanagnou
<janis_papanagnou(a)hotmail.com> wrote:

> mop2 wrote:
>> On Wed, 03 Mar 2010 07:17:18 -0300, LC's No-Spam Newsreading
>> account <nospam(a)mi.iasf.cnr.it> wrote:
>>> Hmmm ... telnet is scriptable at the price of using expect
>>> (which I
>>> had to use only twice in my life, one being the other swith
>>> case). In
>>> the case of the new switches I can't find the relevant info
>>> in the
>>> telnet interface. Also the script is last resort, I am
>>> phasing out
>>> the scripts for interactive use (I'd probably keep them for
>>> crontabs) in favour of java, and prefer to do HMTL parsing in
>>> java
>>>
>> You can try html with shell:
>
> ITYM "bash shell".
>
>> x=0;while read -d\>;do printf "<$x>${REPLY%<*}";x=$[x+1];done
>> [...]


Why?
Well, about the syntax, possibly, but the principle is probably
viable for some others:


$ ksh -c '
> x=0lynx -source http://groups.google.com/not |
> while read -d\>;do printf "<$x>${REPLY%<*}";x=$((x+1));done'
<0>

<1><2>
<3>
<4>404 Not Found<5>
<6><7><8>
<9><10>
<11>
<12>
<13>
<14><15><16>
<17><18>G<19><20>o<21><22>o<23><24>g<25><26>l<27><28>e<29>&nbsp;&nbsp;<30>
<31>&nbsp;<32><33>
<34><35><36><37>Error<38><39><40>
<41><42>&nbsp;<43><44><45>
<46>
<47>Not Found<48>
The requested URL <49>/not<50>was not found on this server.

<51>
<52>
<53><54><55><56><57><58><59>
<60>$
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: Processing stdin in blocks
Next: Bash directory traversal with spaces