On Tue, Jan 23, 2007 at 06:01:01PM +0100, Michal Soltys wrote:
> I've noticed that many programs love to "abuse" TOS byte, lowdelay bit in 
> particular. The effect is, that a lot of packets that one would never 
> expect, land in the 2nd queue (even scp & sftp set it),

NB: He means the second queue mentioned in the filter rule.

My scp and sftp clients (openssh) certainly don't set it.
When I scp, it's ToS 0x8.  When I do interactive ssh, it's 0x10.

However, be aware that interpretation of the ToS byte, now called
the DSCP byte, is totally arbitrary:

http://www.isi.edu/in-notes/rfc2474.txt
http://www.isi.edu/in-notes/rfc3168.txt (ECN bit)
http://www.isi.edu/in-notes/rfc3260.txt (general ISP/NSP info on DSCP)

According to RFC, hosts have to provide a way to configure how
they handle the values of the DSCP byte (called codepoints for
some unfathomable reason).  Since you can hack kernel code, perhaps
you'd like to take a shot at generalizing the current stuff?

Other parties welcome to debate this.

Ostensibly, one would want pf to be able to remark traffic as well;
the general premise is that DSCP is network-specific, and that ISPs
and whatnot would remark at the boundaries.  This seems like a
natural fit for pf, once some syntax and mechanisms are added.
-- 
``Unthinking respect for authority is the greatest enemy of truth.''
-- Albert Einstein -><- <URL:http://www.subspacefield.org/~travis/>

Attachment: pgpoIuhE6zsj7.pgp
Description: PGP signature

Reply via email to