Re: idea: MAC level compression & crypto

Alan Cox (alan@lxorguk.ukuu.org.uk)
Sat, 28 Aug 1999 19:19:22 +0100 (BST)


> I guess, to make such general compression and encryption, that would make
> sense in device queue level.

Note we can't put crypto hooks into a kernel because America is stupid (not
America not Americans). They can go in the kerneli patches. Thats one
reason for just using ssh 8)

> 1. collect - for aironet it would be good to collect about 30-80ms of tcp
> acks and other small packets by which endnodes disturb central node RX.
> For ordinary ethernet too for 10ms - I guess tcp ack of 1/15 maxmtu size
> takes about 1/3 maxmtu packet transmit time when other stations are not
> active.

Packet aggregation works. The scarab driver does this. On a slow link the
aggregation is a win. On a faster link balancing the rx/tx overhead with
the aggregation latency is probably harder to tune but clearly a good idea.

> 2. compress - there was a report on cipe development list on some nice
> short-text-dictionary compressing.

Most packet bodies don't compress. Packet headers can be squashed with care
but the standard VJ compressor assumes packet ordering is roughly maintained

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/