Using FC5 with an Acer Travelmate 230; this laptop has a "winmodem", an Agere AC'97. The Windows driver "Agere Systems AC'97 modem" or AGRSM.sys, connects to my ISP at 41.2Kbps. Did an ATi11 (Agere-specific parameter to report last connection details) and determined Windows is connecting at V.92 w/ V.44 compression.

The readme for slmodem 2.9.9d & 2.9.11 says that it features V.44 & V.42bis compression--However in Linux it connects without fail at approximately 28kbps downstream, 19.2 upstream, using V.34 modulation and V42bis for compression.

Looking through the source I turned on the debug messages in modem.c so they would print to console (that's how I got all those stats on the connection)...But despite slmodem's claim to support V.44 was unable to find hide nor hair of V.44 protocol anywhere in the source; and in fact the places where compression is mentioned seem to assume that compression is either on and V.42bis, or off completely. modem_comp.c seems exclusively written for implementation of V.42, plus numerous references in modem.c where compression state is determined as a binary value, either it's V.42 (which they indicate by setting m.cfg->comp to 3) or it is off.

So does slmodem support V.44 or not? Is it somehow magically handled in that mysterious dsplibs.o or could whoever wrote the readme be lying/mistaken? Noticed that the older slmdm readme claims support for additional compression mode MNP5; what's up with that being dropped in the newer version? Am downloading 2.6.10 kernel source now so I can attempt to compile slmdm 2.7.10 for my 2.6 kernel to see if it works properly. However obviously I am on dialup so have a while to go to download the 44 megs.

Also would appreciate if anyone more familiar with these modems could clarify for me: exactly how much of the modem is "soft"? Is it basically just a fancy sound card that can pick up the phone and all of that beeping and static during negotiation is actually being generated by my CPU? I assume it's got some sort of DSP onboard from the name "dsplibs.o" but is the DSP there to do what? Just FFT type stuff or something more complicated than that? The real low level stuff like whatever generates the DTMF tones and negotiation stuff unfortunately seems missing from the source; presumably those things are in dsplibs.o; should I look in the open-source ALSA module snd_intel8x0m if I want to find that stuff?

---

Found the "AT+MS=?" command in modem_at.c, it reports substantially less modulation types than the Agere driver on Windows. Giving it "AT+MS=34" causes it to connect quicker at the same 28000 with less re-negotiation. Slmodemd's debug messages with AT+MS=34, shows that the "dp" (data pump) immediately tries to go to V.34 mode, whereas without it it goes to "8" then "92" during the first few seconds of negotiation.

AT+MS? shows default for slmodem is V.34, however Windows Agere driver responds to AT+MS? with default setting V.92.

Issuing "AT+MS=92" causes modem to hate me; i.e. lots and lots of renegotiation (as if it's on a crappy phone line; although it's not-- I tried running slmodemd with nice 20 but this actually made it worse) followed by connecting at the same old V.34. I know the +MS command works though because I can give it a slower protocol and get it to connect at 300, 1200, 2400. But if I give it AT+MS=92,0, in an attempt to force connect at V.92 and accept nothing less, it should turn off "auto mode" according to source, but it makes no difference, ends up deciding to connect at V34 anyway. Also tried V.90 which failed connect, and tried hacking dp to use the "K56Flex" option (that protocol is enumerated in modem.c but isn't actually an option via AT+MS), apparently K56Flex was never implemented because that one doesn't even dial out.