20 Kermit Doesn't Work Right with My (RPI) Error-Correcting Modem!
When you buy a modern error-correcting, data-compressing, speed-buffering
modem, you probably expect the modem to perform those functions, and most do.
Unfortunately, some modems that claim to have these features do not have them
at all, but require external software that implements them in your computer,
rather than in the modem where they belong.
One example of this practice is the Rockwell Protocol Interface (RPI), a
proprietary "standard" from Rockwell International Company that allows modem
companies to sell modems at a lower price by incorporating Rockwell chips that
do not include error-correction or compression capabilities.
This "standard" must be licensed from Rockwell, hence you will only find it
implemented in commercial software, such as on the diskette (if any) that came
with your modem.
In general, such software is available only for PCs running Microsoft
Windows 3.x or Windows 95, or built into proprietary communications packages
like Comit.
If your modem documentation says it requires "RPI-compatible" software for
error correction and compression, and you want to use it with Kermit, then you
are out of luck unless you also have the software driver for the modem and can
use it on your computer. Otherwise you bought the wrong modem. Hopefully, you
can return it.
It is usually hard to tell by reading the modem box. One user reports:
"The RPI bit is hidden on the back of the box: '*Error Control* V.42 and MNP
RPI software,' and '*Data Compression* V.42bis RPI software.' The box design
is such that, unless you happen to already know what RPI is, you think you're
getting a modem with MNP/V.42 LAPM/V.42bis compression built in."
If you can't return the modem, you can still use it without error
correction, but then:
- Noise is not filtered out on the modem-to-modem connection, as it would
be with a real error-correcting modem, and noise bursts will interfere with
your online sessions and your file transfers.
- There is no modem-to-modem compression, because that requires error
correction.
- There is no flow control, because that too requires an error-correction
protocol between the modems.
- Speed buffering is ineffective because that requires flow control between
the modems. Thus, if the two modems have different interface speeds, vast
amounts of data will be lost.
Thus, none of the modem's "advanced features" are really there.
Why RPI is a bad idea:
- Implementation of MNP, V.42, and V.42bis in software is a VERY big job.
Since it is uneconomical for software companies to write software for any
platform other than Microsoft Windows, it is unlikely that RPI drivers will
ever be written for DOS, UNIX, OS/2, VMS, or any other platform.
- Even when a company wants to produce a driver for (say) a new
release of Windows, there is generally a long lag time before it is
available. Thus you might find when you install your brand-new operating
system that your modem has become useless.
- The driver software slows down your computer by consuming a vast amount
of CPU cycles over and above what would be used if error-correction and
compression were done in the modem, and it increases memory requirements,
swapping, and in general can be expected to drag down the performance of
your PC.
- RPI seeks to replace an open communication method (the one that is
universally used by serial communication software) by a closed, proprietary,
licensed one, and potentially hold hostage all communications software
developers to nondisclosure agreements.
- It precludes publication of source code.
- Since MNP 2-5 and V.42 and V.42bis are complex protocols, the software
implementations will inevitably be buggy and are unlikely to be consistent,
especially since the "standard" is not an open one, and the implementations
themselves will not be open.
- Even if the drivers are not buggy, the underlying operating system is
likely to be.
- Since not all software in the world will be "upgraded" to "support" the
RPI "standard", your modem will not be usable in many of the ways you might
have expected to use it.
- Many people will buy these modems under the mistaken impression that they
can use their high speeds and advanced features with their favorite
software. The average mass-market consumer is unlikely to understand the
implications of "requires RPI-compliant software" in tiny print on the box.
- By Gresham's Law, "The bad drives out the good". RPI modems are cheaper,
and might well drive reputable modems out of the marketplace, leaving the
entire world's online community with no modems left to choose from but ones
that will work only with Windows. This drives another nail into the coffin
of "legacy" non-Windows platforms, and this, in turn, leaves the public with
fewer choices for operating systems, applications, and computers themselves.
What are the benefits of RPI?
- Lower-cost modems? In order to save a few dollars, you are giving up the
ability to use the modem on the platform of your choice, with the software
of your choice, and you are probably going to get poorer performance than
you would have had with the EC and DC protocols built-in over the life of
the modem.
How do I tell if I have an RPI modem?
Is there a list of RPI modems?
- No. Just about any modem manufacturer is likely to have RPI models. The
modem market is incredibly volatile, fast-moving, and voracious. Any such
list would be obsolete before you could see it. Once-scrupulous companies
will now do anything to cut costs and increase margins. They have no choice
-- if your competitors are doing it, you have to do it too or lose your
business. Rockwell licenses RPI "technology" to anyone who will pay for it,
but is under no obligation to disclose its licensees, nor are the licensees
under any obligation to inform Rockwell (or anybody else) which models
contain RPI chips and which do not. In many cases, the same make and
model can have RPI and non-RPI variations. This makes technical support an
increasingly difficult job. By saving a small amount of money for
themselves, the modem manufacturers have created unneeded confusion among
users and service providers, and driven up costs throughout the online
world.
Is RPI the only "software driven" modem scheme?
- No. There are similar products from other companies. This further
complicates the task of the help desks of the world, for even once they have
ruled out that a modem problem is due to RPI, it might be some other kind of
"controllerless" modem based, for example, on the HSM chipset from AT&T (now
Lucent). US Robotics (now 3Com) sells a "Winmodem" (see next question).
IBM has the Mwave, whose signal processing functions (the very core of a
modem) must be downloaded from the computer.
- We can only expect to see more and more of this in the future. As a
maker of such items commented in the comp.dcom.modems newsgroup, "There's
certainly a market for platform-independent modems, but you can't expect
manufacturers to ignore the market opportunity for Windows-only modems,
especially when they can have a cost and performance advantage."
Of course you can't. But what about end users who don't understand any of
this? Most of them do not make informed choices when buying modems. And
increasingly, many of them make no choice at all -- the modem comes
preinstalled in a PC they have bought, and does not even include a manual;
sometimes not even a brand name. And now that RPI modems have been out for a
while, we are beginning to see how they are passed from hand to hand, installed
in or connected to new computers (for use, e.g., with DOS or Linux) and causing
a whole new wave of problems.
If you have an RPI or other controllerless modem, and you need to use it in
a setting for which a driver is not available -- that is, in most cases, any
platform other than a PC running Microsoft Windows -- you are just plain out of
luck. Return it and buy a real modem. This way, you will encourage modem
manufacturers to continue to make real "platform-independent" modems.
Kermit FAQ / Columbia University / kermit@columbia.edu