kermit.columbia.edu/kermit.columbia.edu.tar/kermit.columbia.edu/newsgroups/misc.20020314-20021006/[e-mail]_Tue Apr 2 11:18:03 EST 2002.msg

Text File | 2002-10-06 | 5KB | 126 lines

Article: 13277 of comp.protocols.kermit.misc
Path: newsmaster.cc.columbia.edu!news.columbia.edu!news-not-for-mail
From: [e-mail] (Frank da Cruz)
Newsgroups: comp.protocols.kermit.misc
Subject: Re: Kermit vs FTP
Date: 2 Apr 2002 11:17:50 -0500
Organization: Columbia University
Lines: 109
Message-ID: <a8clje$5n5$[e-mail]>
References: <[e-mail]>
NNTP-Posting-Host: watsol.cc.columbia.edu
X-Trace: newsmaster.cc.columbia.edu 1017764271 23583 128.59.39.139 (2 Apr 2002 16:17:51 GMT)
X-Complaints-To: [e-mail]
NNTP-Posting-Date: 2 Apr 2002 16:17:51 GMT
Xref: newsmaster.cc.columbia.edu comp.protocols.kermit.misc:13277

In article <[e-mail]>,
Computer <[e-mail]> wrote:
: Hi, I am a university student who downloads many files for my course
: from many ifferent places on the internet and the uni's own servers.
: Now, everyone keeps telling me things such as FTP is quick and Kermit
: is slow or Kermit is good and FTP is garbage, and so forth...
:
: Can anyone tell me what are their main differences (features?) and/or
: what is the same about them to clear things up for me?
:
We haven't written a detailed comparison of FTP and Kermit at the
protocol level, but you can get a fair idea of the functional differences
here:

  http://www.columbia.edu/kermit/iksd.html

and in somewhat more detail here:

  ftp://ftp.isi.edu/in-notes/rfc2839.txt

By the way, the instant you say "ftp", some people might say "Don't
use ftp, use SCP (or SFTP)!"  It's a knee-jerk reaction these
days (similar to "Don't use Telnet, use SSH!"), and it's based on the false
assumption that FTP and Telnet are inherently insecure.  They are not:

  http://www.columbia.edu/kermit/security.html#servers

Of course, neither is Kermit:

  http://www.columbia.edu/kermit/security.html

However, an advantage shared by both FTP and Kermit over SCP and SFTP is
that they are intrinsically platform-neutral, in the spirit of proper
network application protocols.

For eompleteness, I should mention that modern Kermit software does FTP
too:

  http://www.columbia.edu/kermit/ftpclient.html

and for that matter can also act as an SSH service replacing SFTP:

  http://www.columbia.edu/kermit/skermit.html

As for speed, that's a well-worn topic; it's discussed at some length here:

  http://www.columbia.edu/kermit/kermit.html#notslow

FTP has always been fast, because essentially it's no protocol at all -- it
just copies the bytes of the file from source to destination.  Kermit, on
the other hand, is a whole layered communications protocol, itself layered
on top of TCP and IP (or any other transport), originally designed more
with robustness in mind than speed.  By design, the Kermit protocol is also
extensible and has indeed been extended to go fast when conditions allow.

As an aside, in case you think executing a whole protocol stack on top
of another one is silly, you might want to read this:

  http://www.columbia.edu/kermit/case10.html

Modern Kermit programs (i.e. dating from 1998 or so) include a special
adaptation to the Internet called "streaming", which makes their data
transfer throughput about the same as FTP.  Streaming is discussed here:

  http://www.columbia.edu/kermit/ckermit70.html#x4.20

but bear in mind that the benchmarks and figures in that document are based
on equipment and networks that are slow by today's standards.

If you want to compare Kermit and FTP performance yourself, you can do so
easily.  First obtain an up-to-date copy of C-Kermit or Kermit 95:

  http://www.columbia.edu/kermit/

and then use it to transfer files with Columbia's Internet Kermit Service:

  http://www.columbia.edu/kermit/cuiksd.html

A good mix would be a text file, a compressed file (e.g. Zip or Gzip), and
an uncompressed binary file, largish in size.  Transfer the same files
using your FTP client on the same computer on an FTP connection to:

  ftp://kermit.columbia.edu/kermit/

which is the same server, and compare the numbers.  If you have trouble
making an FTP connection, read this:

  http://www.columbia.edu/kermit/ftphlp.html

Obviously there will be statistical variations in your measurements since
the network and servers are not steady-state, so your trials should account
for that.

To make it easy (in case this is a homework assignment), I've set up the
following directory:

  ftp://kermit.columbia.edu/kermit/bench/

It contains three files, each of them about 737K (3/4 MB) in length:

  x.txt  -- Plain ASCII text
  x.bin  -- An uncompressed binary file (an m68k executable)
  x.tgz  -- A tar archive compressed by Gzip

Uploading is problemetic since obviously we don't allow wide-open
unrestricted public uploads, so I'd recommend you measure upload
performance at a site where you have login access.

- Frank