the things people try to sell today...

[Marc Slemko <ma...@worldgate.com>]

http://www.sitra.net/

While I can't dismiss the product because they don't actually say what it
does , I can respond to their "Why the Internet is Slow" document
at http://www.sitara.net/spdseeker/skrprob.htm which says:

    TCP's original purpose is a far cry from the way it is being
    used today; consequently, improvements that
    take into consideration its current use will yield large
    improvements in throughput and latency. For
    example, instead of requesting just the lost packet to be
    resent, TCP resends all of the packets that had
    been sent between the time that the lost packet was sent and
    the discovery that it did not arrive - even if
    all of the other packets arrived at the destination. This
    results in unnecessary traffic on the Internet and,
    in peak times, in high levels of congestion and packet loss.

Hahahahahahahahahahahahahahahahaha.

Funny.  They must have Linux 1.x or _very_ old TCP specs on their
brain, since my world sure ain't like that.  Somehow I don't have
too much confidence in a product coming from a company that claims
they have a understanding of how the Internet works when they say
that.

Re: the things people try to sell today...

[Manoj Kasichainula <ma...@io.com>]

On Sun, Dec 14, 1997 at 12:12:13PM -0700, Marc Slemko wrote:
> http://www.sitra.net/
> 
> While I can't dismiss the product because they don't actually say what it
> does , I can respond to their "Why the Internet is Slow" document
> at http://www.sitara.net/spdseeker/skrprob.htm which says:

[boy do we wish TCP had seletive reject]

> Funny.  They must have Linux 1.x or _very_ old TCP specs on their
> brain, since my world sure ain't like that. 

I wouldn't be surprised if Windoze didn't have selective reject. <g>

> somehow I don't have too much confidence in a product coming from a
> company that claims they have a understanding of how the Internet
> works when they say that.

Well, I've read a few of articles on them over the past few weeks,
which all happen to be linked from their in-the-news page (hey, I'm
not allowed to keep myself up to date by writing free code yet, what
else will I do? <g>) I've heard various different things, such as
eliminating the delay of a TCP connection for each HTTP request
(uhhhh, keepalive?), combining TCP and HTTP (huh??), implementing
selective reject, and something that was muddled incredibly but
sounded vaguely like caching.

My internal BS detectors also go off at full alert from reading their
pages. Maybe this really is useful. Can Bob Metcalfe still be trusted?

Oh, and while I'm posting, an update on my intellectual property
status for the person who is interested: my managers (yes, plural) are
ready to let me write code as I please, but the legal dept. can't
actually give me a written guarantee that Apache wouldn't be hijacked
by an IBM lawyer in a bad mood.

-- 
Manoj Kasichainula - manojk at io dot com - http://www.io.com/~manojk/
"Stocks said 'boo' today..." - Okito Toyoda of NHK on Halloween '96