netdev
[Top] [All Lists]

Re: TCP rto estimation patch

To: weddy@xxxxxxxxxxxx
Subject: Re: TCP rto estimation patch
From: Horst von Brand <vonbrand@xxxxxxxxxxxx>
Date: Fri, 23 Apr 2004 13:35:27 -0400
Cc: netdev@xxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx
In-reply-to: Message from Wesley Eddy <weddy@grc.nasa.gov> of "Fri, 23 Apr 2004 10:24:45 -0400." <20040423142445.GC501@grc.nasa.gov>
Sender: netdev-bounce@xxxxxxxxxxx
Wesley Eddy <weddy@xxxxxxxxxxxx> said:
> The TCP RTO estimation algorithm defined in RFC 2988 is followed properly
> in the kernel, however the constants alpha and beta in the specification
> are hardcoded as 3 and 2 everywhere they occur in the kernel.  Since these
> constants crop up multiple times, this is poor programming practice.  This
> patch binds the numeric values to symbols RTT_ALPHA and RTTVAR_BETA, and
> uses those symbolic values throughout the code.  Since using 2 and 3 for
> these values is not mandatory, this makes tweaking them much easier.

Why RTT_ALPHA and RTTVAR_BETA, and not just RTT_BETA? Or even RTO_xxx?

Is there any reason to change them, ever? What happens if you change them?
Restrictions on values? All this should go with such a patch IMHO (at least
pointers to relevant discussion).

Must go over it with a fine comb to make sure no unrelated 2 or 3 got
replaced... out of my league, sorry.

Your patch is MIME-mangled, =3D, =20, =\n shows up all over the place here.
-- 
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

<Prev in Thread] Current Thread [Next in Thread>