On Wed, 14 Apr 1999, guy keren wrote:
>On Mon, 12 Apr 1999, Stanislav Malyshev a.k.a Frodo wrote:
>
>> I saw such a beast. The most funny thing about it it that is has an
>> advertisement on it, which proclaims that it's "optimized for windows"
>> (beware of the word "optimised for Z", it means meaning is "we are too
>> dumb to make it work with anything but Z"), and that functions are
>> supported by SW instead of HW because.... it's faster and works better! Is
>> it a blatant lie to a customer or just a honest cluelessness?
>
>or to make cheaper modems... if you compare two modems from the same
>manufacturer, where one is a winmodem and the other has hardware
>implementation of the modem's features, you'll see a significant price
>diference for the part of the winmodem. and i don't think this is due to
>economy of scale, but simply due to much cheaper costs (and perhaps also
>because it's a worse device, but for most users today this is not realy a
>problem - their PC has some cycles to spare).
I've been reading a lot of nonsense about why Winmodems are wimpy and
Hayes compatible ones are not. I'd like to dot the few I's I can here:
1. The Hayes standard (AT...) is so badly overwhelmed by modern speeds and
other transmisson standards (voice, fax, voice-under-data etc) that every
Hayes modem implementation since about the first fax/voice modems came
out is a hack, and a major miracle to make work, out of the engineering
point of view. YOU pay for that, and YOU live with the bugs.
2. The so called Hayes 'standard' has about 5 standard commands, and the
rest of them, look like a jungle, only worse, with each manufacturer
having a go at a few parameters here and there. Ever wondered why so many
modems are hard to get working in auto answer and fax mode with non-OEM
software ?
3. The concept of the Winmodem is coming from somewhere else, and it is
more advanced than the Hayes modem concept. Moreover, with the current
speed of evolution of standards (56k, 64k etc) it is a major hole in the
manufacturer's pocket to maintain the chipsets and firmware. Note that ALL
flash 'upgradeable' modems qualify as type 2 winmodems, they only happen
to run Hayes type firmware for your convenience (and dollars. Especially
dollars. Flash memory costs a lot of money, and so does the toolchain to
keep fresh firmware downloadable etc. A DSP based winmodem is just the
same thing - but it does not have flash - so it's a wimpy winmodem in
your eyes... ).
4. The 'Linux can't use Winmodems' myth is due SOLELY to the fact that
winmodem makers do not bother to communicate the register level interface
of their products to Socialist users (and more important, to the prying
eyes of the competition, always out for a new idea or two). There is no
problem whatsoever to write a winmodem driver for Linux, for a given
winmodem type, once the spec is available.
5. All Winmodems are not made equal: There are those with a coprocessor
and those, without. All high speed winmodems I know of DO have a
coprocessor. So much for 'wasting cycles' of the PC. Winmodems without a
coprocessor are basically a proprietary D/A and A/D each with a FIFO in
front of it and some periphery that can do DTMF (dial tones) and control
the off hook relay. Winmodems with a coprocessor are as good as, or better
than, Hayes compatible modems. They contain a RISC DSP processor that does
most of the data processing running a program loaded into it by the PC
itself, when it opens the device (and sometimes, several times while it
works). FYI the bt848 video capture card does the exact same thing (it has
a RISC coprocessor). It does not even have its own memory, it uses the
system's through the PCI bus. It costs $80 (the full TV PCI card, that
is), less than many a modem. Wonder why ? This is the way to go it seems.
6. Out of the point of view of Linux and FreeBSD etc users, winmodems are
anatema. But it would be easy for the manufacturer to supply a binary-only
driver for his device, on a beta or alpha basis, and make several thousand
potential users happy. This hasn't happened yet, but maybe... For example,
I believe that the bttv (bt848) data sheet, which was available on the
net, and good to download by anyone, has increased sales of bt848 based
cards by several percent. I know quite a few webcam sites that use this...
and this did not lower the price at all !
It is good for hardware makers to publish a register level interface to
their products, because they can sell more. Linuxers, engineers, hobbyists
and others will prefer their products, and they will have a longer shelf
life like this. I think that many people notice this already, there are
(mostly unofficial) Linux and other drivers for various hardware floating
around, mostly coming from the makers (unofficially) etc.
sorry for the long post, I had this on my heart for a while now,
Peter
