[1522] in Commercialization & Privatization of the Internet
Re: Alternate Routing
daemon@ATHENA.MIT.EDU (Manavendra K. Thakur)
Mon Oct 21 23:07:26 1991
To: Vince Fuller <vaf@Valinor.Stanford.EDU>
Cc: sommerfeld@apollo.com, yakov@watson.ibm.com, steve@ncri.cise.nsf.gov,
In-Reply-To: Your message of Mon, 21 Oct 91 19:20:18 -0700.
Date: Mon, 21 Oct 91 23:07:09 EDT
From: "Manavendra K. Thakur" <thakur@zerkalo.harvard.edu>
>>>>> On Mon, 21 Oct 91 19:20:18 PDT, Vince Fuller <vaf@Valinor.Stanford.EDU> said:
> [Vince quotes from Bill Sommerfeld <sommerfeld@apollo.com>:
> If you *really* want to use ToS to distinguish between
> "commercial" and "noncommercial" traffic, you're really going to
> have to be able to get *routers*, probably on the boundary
> between the mid-levels and campus nets, to fill in the
> "appropriate" ToS bit in the IP header while the packets are
> passing through.
> The only entity which can properly set the TOS is the user. The
> "easy" solution to this problem is to use a special TOS bit (or
> combination) to indicate "commercial/research" and modify each
> application to set the bit appropriately based on user input.
> Again, the "easy" way to accomplish this is to make a special
> "commercial" version of each application that sets the appropriate
> TOS bit (i.e. "ctelnet", "cftp", etc.). Mail would be a special
> case, since its delivery is not usually handled by the user agent -
> some sort of header scheme would need to exist. I say "easy" in
> quotes since doing this creates a user education problem, but we
> really already have that problem for sites which can originate both
> research and commercial traffic.
>
> --Vince
Vince,
You *are* kidding, I hope. Do you seriously recommend that we double
the amount of disk space needed to hold all these duplicate
applications, and go to all the effort to modify all those mail
programs and telnet implementations from supercomputers on down to PCs
-- just to add support for one more bit?!
Forget about the user education problem, what about the implementation
problems? That problem doesn't sound easy to me. Sounds more sort of
like crazy.
If we're really going to have to go through all that work, then let's
at least get more bang for our effort. If we really want to take on
the herculean task of modifying all those mail programs etc, it would
make sense to do so only *after* we have a well-defined description of
an overall ToS (or QoS) scheme that can do more than just distinguish
between commercial and R&E traffic.
Clearly, formulating such a well-defined description is what standards
bodies are supposed to do. Or if you don't like standards bodies,
then give it to the IETF, or the Berkeley unix crowd, or somebody else
to your liking.
In the meantime, I think Bill's proposal is much more workable and
palatable, IFF we have to do anything at all. Here's why I feel this
way:
It has been repeated over and over on the com-priv mailing list that
users don't care about networks or the protocols used on them.
Rather, users care about applications that can take advantage of the
network's capabilities.
Now given that implementing a means to distinguish between commercial
traffic and R&E traffic is a network issue in the purest sense, I
don't think it makes any sense to force users to decide whether to use
"cftp" or "ftp" (to use your model).
Rather, Bill's proposal lays the implementation and operational burden
where it properly lies: on the router vendors and network service
providers, whose whole purpose in life is, needless to say, to provide
and connect networks.
Bill's proposal would also place the burden squarely on the shoulders
of NSF and other policy making bodies (e.g. the regionals) to define
exactly the criteria by which a site would be designated as commercial
or R&E.
The users won't have to be involved, and no applications need to be
rewritten - as is proper.
Again, all of this applies IFF we (the collective we) actually have to
do something. Frankly, I think this whole concept of commercial vs
R&E traffic is a cockamamie idea through and through, and I wish it
would just go away so that we could deal with more important questions
and problems. Sigh.....
Manavendra K. Thakur Internet: thakur@zerkalo.harvard.edu
Systems Programmer, High Energy Division BITNET: thakur@cfa.BITNET
Harvard-Smithsonian Center for DECNET: CFA::thakur
Astrophysics UUCP: ...!uunet!mit-eddie!thakur
