Hi Avi,

> I was under the impression that it goes inside -> out
Source to Sink. I find it is easier to visualize when each filter is
on its own line with appropriate indentation.

> I think you mixed encrypt with Gunzip instead of Gzip..
Perhaps - I copied/pasted from a few different projects. I did not run
the example (I know not testing it is usually a bad idea).

> (A)
> 1. GZip 1024;
> 2. Encrypt 1024 zipped bytes;
> 3. Store intermediate result in sink?
This is correct. Compression develops entropy in the plain text before
encryption.

> Same goes for block ciphers - won't I need to have (at least) the last
> block from the previous iteration for XORing / other purposes - with
> the next block?
No. When you pair a Mode/Block Cipher with a StreamTransormation, the
details are handled for you. See
http://www.codeproject.com/KB/security/BlockCiphers.aspx.

Jeff

On 2/9/09, Avi <[email protected]> wrote:
>
> Hey Jeff,
>
> > When using it in pipelining, you will want the compressor outermost so
> > that it process data first. Below, data travels from TOP/LEFT to
> > CENTER/RIGHT.
>
> A very important observation!
>
> I was under the impression that it goes inside -> out, like it usually
> does, e.g.:
> MeLast( MeNext( MeFirst() ) );
>
> --
>
> For general knowledge: in the example you gave (I think you mixed
> encrypt with Gunzip instead of Gzip.. no biggie), let's say I have a
> source file of a few kilobytes, and let's say PumpAll() pumps 1024 on
> each iteration. Is the behavior going to be:
> (A)
> 1. GZip 1024;
> 2. Encrypt 1024 zipped bytes;
> 3. Store intermediate result in sink?
>
> Or is it going to be:
> (B)
> 1. Gzip the entire stream;
> 2. Encrypt the entire zipped result;
> 3. Store the entire encrypted result in sink?
>
> I tend to believe that it'll be (A), but here's where I'm lacking
> knowledge again - won't I need to scan all the input bytes in order to
> produce a better compression scheme (dictionaries, etc) than zipping
> unrelated byte portions?
> Same goes for block ciphers - won't I need to have (at least) the last
> block from the previous iteration for XORing / other purposes - with
> the next block?
>
> Thanks,
>
>    Avi.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the "Crypto++ Users" 
Google Group.
To unsubscribe, send an email to [email protected].
More information about Crypto++ and this group is available at 
http://www.cryptopp.com.
-~----------~----~----~----~------~----~------~--~---

Reply via email to