[plt-scheme] no #!eof

Only by access to the next lower level (device being the lowest, then 
blocks, rbts, fats etc and at the top the stream itself).But the nice thing 
of a stream is that it abstracts all lower levels away. Therefore it should 
be imposible to continue writing after the stream has been declared to be 
ended, Therefore it should not allow *writing* an eof mark. At the stream 
level eof must not be a datum. Hitting the end of a stream is a kind of 
exception, although not necessarily an erroneous one. Oh yes, the eof-object 
is a datum, but fortunately writing this datum does not write an eof mark!

How do we know that we hit the end of a proper list? Because nil is not a 
pair and because Scheme makes it impossible to replace nil by a pair (which 
is quite different from replacing the last cdr of a mutable list, for this 
acts on a pair, not on nil) Imagine a Scheme that allows taking the cdr of 
nil. The whole language would collapse into one big contradiction. When we 
take the cdr of nil, we don't interpret the resulting exception as a datum, 
do we? Allowing to read on after hitting the end of a stream is like making 
cdr return the exception as a value rather than passing the exception to the 
exception handler.

You may also compare this to reading a damaged magnetic tape (in the old 
days). In order to do this you have to read the tape at a lower level, 
closer to the hardware. You'll have to interpret the contents of individual 
blocks, (both label and data blocks) yourself and in some cases you can only 
know that there realy is no more to read after the system signals that the 
end of the physical tape has come loose from the reel (angry operator at the 
main frame console, for he/she has to come out of his/her cosy and luxery 
chair and it may require some effort to put the tape back into the reel) And 
even if the tape come loose from the reel, you cannot be sure that you have 
read it all for the tape may be a volume of a multi volume tape, which you 
cannot know for sure in case volume information has been damaged) On a 
random access device things are even far more complicated.
mho, Jos

----- Original Message ----- 
From: "Joe Marshall" <jmarshall at alum.mit.edu>
To: "Abdulaziz Ghuloum" <aghuloum at cs.indiana.edu>
Cc: "Dave Herman" <dherman at ccs.neu.edu>; "PLT Scheme Mailing List" 
<plt-scheme at list.cs.brown.edu>
Sent: Tuesday, April 15, 2008 9:46 PM
Subject: Re: [plt-scheme] no #!eof


> If you can put EOF markers in the middle of a stream, then how do you
> reliably detect if there is more to be read?
>
>
> -- 
> ~jrm
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-scheme