home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
kermit.columbia.edu
/
kermit.columbia.edu.tar
/
kermit.columbia.edu
/
newsgroups
/
misc.20021006-20030409
/
000028_fdc@columbia.edu_Sun Oct 20 13:10:39 EDT 2002.msg
< prev
next >
Wrap
Text File
|
2003-04-08
|
2KB
|
46 lines
Article: 13791 of comp.protocols.kermit.misc
Path: newsmaster.cc.columbia.edu!news.columbia.edu!news-not-for-mail
From: fdc@columbia.edu (Frank da Cruz)
Newsgroups: comp.protocols.kermit.misc
Subject: Re: Problems transferring over ppp connection
Date: 20 Oct 2002 13:10:36 -0400
Organization: Columbia University
Lines: 29
Message-ID: <aouo2c$rc0$1@watsol.cc.columbia.edu>
References: <1034801647.495773@irys.nyx.net> <1034909930.793290@irys.nyx.net> <m1n0pab9d2.gnus@usa.net> <1035092921.407864@irys.nyx.net>
NNTP-Posting-Host: watsol.cc.columbia.edu
X-Trace: newsmaster.cc.columbia.edu 1035133837 18893 128.59.39.139 (20 Oct 2002 17:10:37 GMT)
X-Complaints-To: postmaster@columbia.edu
NNTP-Posting-Date: 20 Oct 2002 17:10:37 GMT
Xref: newsmaster.cc.columbia.edu comp.protocols.kermit.misc:13791
In article <1035092921.407864@irys.nyx.net>,
Henry van Cleef <hvanclee@nyx.net> wrote:
(More about Kermit vs Netscape file transfers over a PPP connection.)
When Kermit makes its own dialup connection, it controls the serial-port
and modem settings, the i/o, flow control, and buffering. When Kermit
uses a PPP connection, it has no control over these things, and in fact,
has no way of knowing whether the underlying TCP/IP connection is serial,
Ethernet, wireless, or what.
The TCP layer presents a uniform interface to applications that shields
them from the nature of the underlying connection. It is the job of the
lower layers (TCP, IP, PPP, the serial port, and the modem in this case)
to provide the application with the highest quality connection.
Of course we realize this doesn't always happen, which is why Kermit has
so many user-level adjustments: streaming vs Ack/Nak; packet length,
window size, various adaptations to lack of transparency. Kermit can't
see though the TCP layer to get at the datalink layer or physical devices.
If they are not doing their jobs, then you have to step in and either fix
them or else scale back on Kermit's performance settings.
As to why Netscape transfers work and Kermit ones don't, I'm sure this
could be pinned down given enough evidence, such as TCP packet traces.
Once you see the characteristics of the HTTP data transfer, you could
adjust Kermit to the same characteristics and it would work the same way.
- Frank