Linux TCP/IP stack.

Gregory Maxwell (linker@z.ml.org)
Sun, 11 Jul 1999 15:45:04 -0400 (EDT)


Hi, in recent threads on various amiga lists, there were made some serious
claims about Linuxes TCP/IP stack. Basically they said that it was worse
then MSes (as far as standards complience) and that it lacked many
important new features, such as QOS, multicast, IPv6 and such..

This struck me as crazy talk, but it's backed up by the author of the
Miami Amiga TCP/IP stack.. But lots of people are saying it now, and I'm
sure that some goofball in the press is going to see and and take it as
the truth..

Can anyone comment on this:

http://www.flyingmice.com/squid/moobunny/amiga/messages/10172.shtml

>"A big problem is that the TCP/IP code in Linux is not BSD-derived, but
>adapted from some other experimental third-party implementation. Rumor
>has it Linus misinterpreted the BSD licensing conditions and decided not
>to use their code because of that.
>
> As a result the TCP/IP code in Linux is by far the worst in the industry,
>worse even than the severely broken code in Win-NT/98. Its performance
>does not scale well, in particular under congestion, and bugs in the
>kernel are holding back the deployment of new Internet protocols because
>of interoperability problems, and are curbing performance on the Internet.
>See recent RFCs on "Known TCP implementation problems". They don't mention
>
> Linux there explicitly, but many of the problems listed ARE caused by
>Linux. Miami contains special code in four places only to work around
>known bugs in Linux and maintain interoperability. Add to that the
various bugs in the Linux TCP/IP code typically caused by race conditions
>or unhandled errors (things like TCP connections remaining in CLOSE_WAIT
>indefinitely). "

and

http://x38.deja.com/[ST_rn=if]/threadmsg_if.xp?thitnum=15&AN=499729807.1&mhitnum=0&CONTEXT=931720044.2066284594

-
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/