On Fri, 30 Apr 1999, Leszek A. Szczepanowski wrote:
> On 30 Apr 1999, Jonathan NAYLOR wrote:
>
> > I still firmly believe that compression should be on a per application basis,
> > and should not be built into a transport protocol.
>
> So programmers, implement zlib compression for pop and smtp
> daemons and this should be end of this discussion. For
> compatibility also LZW should be implemented. This is
> what we all waiting for so long time!
I would disagree, and say it should be a fair bit lower than the
application layer on the OSI model.
The wonderful thing about using linux for amateur radio is that there are
already excellent smtp, pop, and every other type of TCP/IP daemon. If you
were to approach the developers of such daemons, and ask them to support
compression, I suspect you'll not get very far. The benefit just isn't
there for them.
The reason we need encryption is because of the slower speeds at which
amateur packet radio generally takes place. While there is fast packet
radio, that's definately the exception, and not the rule. Compression can
actually be a bad idea on some networks. Trying to do compression on a
switched fast ethernet, like I have at work, would likely just slow down
traffic and send my CPU use through the roof.
If you look at where you see compression, it's often in the lower levels
of the OSI model. My ISDN adapter does compression, transparent to the end
parties. You'll often find compression on slow serial links. You'll find
it in tunneling protocols, like Microsoft's PPTP. Only occassionally do
you find it at the application level, like with ssh. Even ssh is something
of a special case, given that it acts both as application and as TCP
tunnel.....
Bob N2KGO
