cat'ing a binary messes up terminal - why?

[Ben Petering]

David Lloyd wrote:
>
> Ben,
>
>> I have a query, not important, I'm just curious. Hopefully someone
>> understands what's happening here.
>>
>> If I say
>> $ cat /bin/login
>>
>> for example, my terminal gets all messed up - i.e. typing '/bin/login'
>> now appears as
>>
>> $ /␉␋┼/┌⎺±␋┼
>>
>> I suppose some kind of character table translation is broken or
>> something... can someone explain in detail why this happens? I NEED to
>> know :)
>
> The terminal accepts a stream of bits which it may interpret in
> different ways, for example:
>
>  1. It may interpret a sequence as an instruction to display the letter
>     "A"
>  2. It may interpret certain sequences in different ways, for example,
>     when you press "ctrl + c" (ctrl^c) a certain set of bits is sent to
>     the terminal and that sends a "TERM" signal [generally] to the
>     currently running process in the terminal
>
> Now, different terminals have different capabilities; see:
>
>  1. http://www.tldp.org/HOWTO/Text-Terminal-HOWTO-15.html
>  2. man terminfo
>
>
> Some terminals are obviously able to display "unicode" and other
> character sets.
>
> Essentially, what I'm saying, is that a terminal will attempt to print
> out a sequence of bits as sensibly as it possibly can.
>
> Now, obviously terminals can't really accept "random" sequences of
> bits. There needs to be some standard way to send sequences of bits to
> the terminal to get it to do something. This is called an "encoding"
> and yes, there is more than one possible encoding.
>
> When you "cat" a binary file, though, there is no "sensible" encoding.
> Consequently, the terminal tries to do its best and you could get
> absolutely anything (including a sequence of characters or escapes
> that you don't want).
>
> Hence, why you see garbage.
>
> If you do see garbage, you should type:
>
>  reset
>
> ...and most terminals will set themselves back into a known state. It
> is possible to send terminals sequences of bits that change encoding
> and/or do other weird things and you get weird results.
>
> Consequently, the short answer is:
>
> "It's showing you weird characters and making strange noises because
> the sequence of characters it is trying to display is not encoded in a
> way to display what you are sending it. It's the Garbage In Garbage
> Out principle.".
>
> Why do you need to know?
>
> DSL
>
>
>
>
Thanks David, Thomas, Karl. Very informative.

Incidentally, I first encountered this years ago when I was (more of) a
newbie. Back then I also discovered that catting the same file and
interrupting the output at arbitrary points would 'reset' the terminal,
and I could see what I was typing again. But of course, 'reset' is the
proper way. (Thanks David!)

> There are even some exploits based on specially crafted binary files, so
> be careful  :-) 
> "less" is a safer choice than "cat", "hexdump" is safer still.

Thomas, I like the way you think. :) This comment reminds me of Georgi
Guninski's web page (http://www.guninski.com/): 'Best viewed with a
sniffer'.

Just for the record, I don't make a practice of downloading possible
malware and catting it. Honest. :) And if /bin/login has been modified
maliciously then I've got bigger things to worry about than specially
crafted binary files designed to break terminals. Hehe

BTW David, I 'need' to know because I am compulsively curious. That's all.

Ben