Re: [security-services] Bug in the bindings doc; wrong RFC

[Greg Whitehead <grw@trustgenix.com>]

On Feb 8, 2005, at 5:21 PM, Scott Cantor wrote:

>> So, the first observation is that it really wouldn't be that heinous 
>> to
>> use rfc1951, if we wanted to do that -- as long as we restricted CM to
>> 8 (deflate). I still think we were correct in specifying rfc1950,
>> however, so let's do some experiments...
>
> We specified 1951, not 1950.
> 1951 is DEFLATE and 1950 is zlib.

Right. Sorry, for the confusion.

>> In the first case, the 'nowrap' argument to the Deflater is false (so
>> the ZLIB header and footer are added). In the second case, the 
>> 'nowrap'
>> argument is true, so we just get the deflated data. Note that I was
>> deflating zero-length data in these tests, so all we are seeing are 
>> the
>> headers etc.
>
> Did you understand the thing in there about the dummy bit required with
> nowrap? That's what had me worried. Otherwise it's just a simple case 
> of the
> class names being a bit off, since the javadocs do say that it's for 
> zlib.

Yeah, it's a retarded API if you ask me, as they could have done this 
internally. Basically, you have to feed it one additional byte of 
input, beyond what you want to decode, before it will tell you that 
it's done... it appears to be doing some lookahead thing that needs to 
look beyond the end of the input. Brilliant.

-Greg