[going offtopic] Re: my broken TCP is faster on broken networks

Andrea Arcangeli (andrea@e-mind.com)
Fri, 11 Sep 1998 19:04:13 +0200 (CEST)


On Fri, 11 Sep 1998, Alan Cox wrote:

>The maths is clear. And if your ISP is slow and you get heavy packet
>loss and long delays from many sites (ie its not their end) then
>change ISP. Vote with your feet.

The _only_ irregular money I get are from my ISP and so I am very happy
with it ;-) (note also that all _other_ ISP in my city don' t know the
word UNIX (eventually they know the word "SCO" (don' t ask me why) but
sure they don' t know Linux and [heheheh] Solaris) and they all have M$
hats on the head). My ISP pays really a lot of money for its internet
connection (more that I could guess, really), they are nice guys. The
point is that nobody except 1% of users will care about telnet
responsiveness. http is fast everywhere (also connecting to my
University). It _seems_ that all routers near me drop every packet except
if it' s source or destination port is 80 8080 or 443 (I' ll try to put a
ssh daemon on port 80 soon ;-). Maybe the telnet slowdown is only due the
fact that telnet does a lot of pushes and send very small packets. Maybe
the collapsed routers should only decrease their m[tr]u and start
fragmenting packets to scale better (but I am _sure_ that 99% of firewall
admin where I live don' t know how to do that).

I' m afraid to say this but to give more sense to your suggestion you
should change it in:

"... then change your City.", or better "... then change your State.",

because here your first proposed solution doesn' t work.

The only good point is that here I can test very well the TCP retransmit
code ;-).

Andrea[s] Arcangeli

-
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/faq.html