This history is largely excerpted from the Linmodem HOWTO at
http://walbran.org/sean/linux/ and exchanges of  
discuss@linmodems.org participants, with Archive access at 
http://www.linmodems.org

The first port of Lucent chip support to Linux was done by Eric Bentley,
as requested by Red Hat, to provide  support under kernel 2.2.12  kernels. 
This was ltmodem.o driver, version 5.68.2.2.12. An interesting commentary 
on support issues is available at http://lwn.net/1999/1209/a/lucent.html 

This binary sufficed into kernel verions 2.2.15, but required 
forced (-f) insertion:
   insmod -f ltmodem
to bypass the module-kernel version mis-matches.  Mark Spieth provided a 
fixscript tool that masks such mis-matches (available at 
http://www.test.dclabs.com.au/linmodem/fixscript) so that there sufficed a 
simpler:
   insmod ltmodem 

Source code changes within the 2.2.15 series faulted use of the ltmodem.o 
binary. While dail up and login could be achieved, the ppp protocol could no 
be established.  There was frequently a system crash when it was attempted.
Christoph Hebeisen experimented and discovered a bypass. Provided that ppp.o 
compiled from 2.2.14 sources was used, the ppp protocol could be established:
   insmod ltmodem
   insmod slhc
   insmod -f ppp (compiled under 2.2.14)
This trick served into kernel versions 2.2.16, though with increasing 
flakiness as kernel versions increased.  David Hinds fixed a bug, which caused 
ltmodem.o to grab other serial ports.  

These tricks also proved to be effective for modems with a ESS chipset, as 
worked through by Willie Green.  To more efficiently convey these technicalites
to the increasing population afflicted with winmodems, Sean Walbran originated
the Linmodem HOWTO, with contributions from Marv Stodolsky.
 
Meanwhile, Mark Spieth searched the source code for the change breaking ppp 
protocol support.  It was recognized within include/linux/tty.h of the 2.2.16 
source code tree. Mark provided a fix/patch.  Marv Stodolsky reported a few 
hours later, that the same fix was effective under 2.2.17.  With these tty.h 
fixes, kernel version matched ppp.o could again be used. Reliable ppp service 
enused, but with worries about functionality under the maturing 2.4.nn kernel 
series.

Then there was an exciting week, coinciding with the Thanksgiving holiday in 
the USA.  Jean Gomes reported the presence of a compiler kit at 
http://www.lectron.com.tw/Linux.htm.  It was provided by Lucent/Agere for 
 major customeres, initially suited for service with the  2.2.14 kernels.
It includes a  proprietary binary, ltmdmobj.o , and open source components. 
These components can be linked and compiled with kernel sources, 
producing ltmodem.o driver 5.78.  This ended the Dark Ages of newer kernels 
and version mis-matched ltmodem.o.

Within a day Christoph Hebeisen and Mark Spieth improved the Makefile to 
support compiles with sources upto 2.2.18. A day later, Marco Giordani 
lofted us into the realm of maturing kernel 2.4 sources, providing patches 
enabling compiles with kernel-2.4.0-test11 sources. Mark Spieth has and is 
continuing successive improvents. They are available as ltmodem5.78*.tar.gz
kits for public download at http://walbran.org/sean/linux/stodolsk/.

During December, 2000, Jacques Goldberg was trying to implement service of 
ltmodem.o on a new PC.  He found that it had a default PNP (Plug and Play)
setting "hidden" in the BIOS Advanced menu.  As soon as a change from PNP 
to OTHER was made, good ltmodem.o service enused.  Many other Users have 
since been helped by recognition of this problem. It is expected/hoped that 
it will be largely obviated with the transition to PNP competent 2.4.0 kernels,
but some problems have been reported to the List.

A problem was encountered with the Xircom modems included in Toshiba Satellite 
laptops.  They appeared to have the MARS chip characteristic of the Lucent 
winmodems, but did not support insertion of ltmodem.o.  With advice from 
Javier Nigro, Andreas Plesch changed a PNP device-ID code within ltmodem.o, 
and full functionality ensued.  

Vladimir Chernyshov identified and fixed a problem under 2.4 kernels,
causing ltmodem.o to raise a serial.o driver with a shared IRQ, This fix
brought substantial improvements of performance under 2.4.nn kernels.  
Marv Stodolsky reported that while the modem could be accessed through 
several IRQs, most did not support dial tone acquisition.  This problem
NO DIAL TONE problem also afflicts other winmodems. 

The current ltmodem5.78e.tar.gz kit from Mark Spieth has transparent support 
for a broader range of Xircom modems.   Peter Fox provided patches
supporting use of his serial monitor tool: http://www.roestock.demon.co.uk/serialmon/
Mark has implemented the rpm installers.  Marv Stodolsky updates some of 
the documentation.

Russell Coker identified and implemented a fix for devfs under 2.4 kernels,
changing the name to ttys/L0 if enabled and disabling the callout device
registration.
 
Lastly, without the usefull exchanges with MANY other participants of 
discuss@linmodems.org, the rapid progress over these past few months would 
certainly not have been achieved.  To avoid email address capture by SPAM 
organizations, personal contacts are not herein given.  But any queries 
will surely be responded to by the discuss@linmodems.org Collaboratory


