[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Tos in IP header

On Tue, 19 Feb 2002, Marc REYNES wrote:
> i'm looking for a method to evaluate the gain in performance involved
> by the sets of TOS bits.

There is no performance gain, in general.  You can configure a Linux
system to use the TOS bits as routing keys and, thus, select the
outbound path based on them.

This is very seldom useful, however, and not really worth the bother.

It's worth noting that almost /none/ of the widely used hardware,
notably all Unix variants, the Cisco hardware and Windows, don't respect
the TOS bits by default and, in many cases, don't support using the TOS
at all.

The situation is reportedly the same with other systems, but I have not
verified this myself.


> I've some extra question : 
>  - this exemple is from the official ferm's homepage, the author
>  prevent from using ToS for UDP, can someone explain me why ?

TOS is a TCP concept, not an IP concept. This implies that all protocols
except TCP, such as UDP, IGMP, ICMP, etc, have not got those bits.[1]

As such, you /can't/ set the TOS bits on a UDP packet.

>  - why it is recommand to set only one bit ? (from ipchains man)

Because you are even less likely to get the service you want with more
than one bit set and because the RFC defining them suggests that this is
the best way to approach the situation.


[1]  I believe that, of the well defined protocols today, only TCP and
     SCTP define anything along these lines.

Specialized meaninglessness has come to be regarded,
in some circles, as a kind of hallmark of true science.
        -- Aldous Huxley

Reply to: