----- Original Message -----
From: "Enrique Arredondo" <atk@sbcglobal.net>
Newsgroups: comp.unix.sco.misc
To: <distro@jpr.com>
Sent: Thursday, May 08, 2008 6:50 PM
Subject: Re: OSR 507 using an external modem with an USB serial to Ethernet adapter


>
> "Enrique Arredondo" <atk@sbcglobal.net> wrote in message
> news:fH1Uj.15927$2g1.4046@nlpi068.nbdc.sbc.com...
>>
>> "Brian K. White" <brian@aljex.com> wrote in message
>> news:029701c8ab4b$77248c80$6b00000a@venti...
>>
>> ----- Original Message -----
>> From: "Enrique Arredondo" <atk@sbcglobal.net>
>> Newsgroups: comp.unix.sco.misc
>> To: <distro@jpr.com>
>> Sent: Tuesday, April 29, 2008 5:25 PM
>> Subject: Re: OSR 507 using an external modem with an USB serial to
>> Ethernet adapter
>>
>>
>>>
>>> "Brian K. White" <brian@aljex.com> wrote in message
>>> news:036901c8a979$f60c7e50$3a8d53d0@venti...
>>>
>>> ----- Original Message -----
>>> From: "Enrique Arredondo" <atk@sbcglobal.net>
>>> Newsgroups: comp.unix.sco.misc
>>> To: <distro@jpr.com>
>>> Sent: Monday, April 28, 2008 3:52 PM
>>> Subject: OSR 507 using an external modem with an USB serial to Ethernet
>>> adapter
>>>
>>>
>>>>I have a device called the UDS1100 made by lantronix that let's me
>>>>convert
>>>> any serial device into a TCP IP device, I tried this on printers and
>>>> dumb
>>>> terminals and it works really good. My final task is to move my fax
>>>> modems
>>>> that I run with VSIFAX on unix from tty1A13 or tty3E13 into this new
>>>> device.
>>>>
>>>> Has anyone played out with this configuration and figure out a way of
>>>> making
>>>> the scripts or dialers that communicate with the modem to hit the
>>>> assigned
>>>> IP address and port so the modem works ?
>>>>
>>>> In the case of the printer my "model" file starts with :
>>>>
>>>> #!/bin/sh
>>>> YOUR_PRINTER_OR_IP="192.168.2.19"
>>>> PORT=9100
>>>>
>>>>
>>>> and then I just assign the printer to that model and use /dev/null and
>>>> that
>>>> works out great.
>>>>
>>>>
>>>> Can I do the same trick for the external modem thru the
>>>> "Serial-Ethernet"
>>>> device that wants to be called on port 10001 ?
>>>> IP address= "192.168.6.182"
>>>> Port="10001"
>>>>
>>>> Thanks
>>>
>>> I have used a few different serial device servers (digi, equinox, perle)
>>> with fax modems and both hylafax and vsifax.
>>>
>>> However:
>>>
>>> * In all cases the device had a driver and/or daemon that provided tty
>>> device nodes that work just like ordinary hardware serial ports. vsifax
>>> didn't need any special config, just tell it the tty device name and
>>> perhaps
>>> doctor the permissions on those tty's for vsifax.
>>>
>>> * You are really not supposed to use fax modems over tcp like that
>>> because
>>> the fax protocol is very timing dependant, and the tcp/ip layer makes few
>>> promises about timing, so you may or may not see a lot of failed faxes
>>> even
>>> though everything is configured absolutely perfectly. This is not my
>>> opinion
>>> but the word right from ifax (hylafax) and VSI.
>>>
>>> I had a lightly loaded gigabit network and only about 8 modems on a 32
>>> port
>>> digi (or a 16 port equinox for a while). Nics, switch, and digi were all
>>> pretty over specced for the dribble of traffic 8 14.4kb modems can
>>> deliver.
>>> It _mostly_ worked ok. I don't do that any more though.
>>>
>>> --
>>> Brian K. White brian@aljex.com http://www.myspace.com/KEYofR
>>> +++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++.
>>> filePro BBx Linux SCO FreeBSD #callahans Satriani Filk!
>>>
>>>
>>> How about if I create a "VIRTUAL Serial Port" and then use that one
>>> instead
>>> ? Is there such thing on SCO Unix ?
>>
>> What did I just say?
>> Thats exactly what I just said the drivers for most such devices provide.
>>
>> I looks and smells just like a serial port, so, from the software's point
>> of view it thinks it's fine.
>>
>> But that doesn't mean it will actually _work_ well, even though the
>> virtual port driver is perfectly robust.
>>
>> The problem has to do with timing. The fax protocol requires things to
>> happen within specific windows of time, and tcp/ip does not garantee that
>> any given packet will get delivered within any specific window of time.
>>
>> Confusingly, on most reasonable lan's it'll actually work ok mostly, but
>> even on an over-spec'd lan, tcp/ip just plain doesn't make promises about
>> how long a given packet will take to get delivered. That means sometimes
>> the fax protocol will break no matter what. Even on a lightly loaded lan
>> there are spikes and bursts of other traffic, and those will cause your
>> virtual port traffic to incur little pauses, a packet may take 300 ms to
>> get delivered instead of .01 ms, and the fax that was in progress fails
>> because the handshaking protocol broke because the server needed to
>> acknowledge the next phase within 50ms and it took 10 ms to actually do
>> the work and then the virtual port took another 300 ms to react to a
>> command because the tcp/ip packet that delivered the command took that
>> long to get to the serial device server.
>>
>> If you had a switch that allows you to set QoS levels then you could give
>> traffic on the tcp port that the virtual serial driver uses higher
>> priority, like what you do for voip. That would probably reduce the
>> occurance of delayed commands significantly.
>>
>> And if you use only fax class 2 modems then since they do most of the fax
>> protocol in the modem hardware, that will probably also help to ensure it
>> works even more often since it would reduce the number of timing-critical
>> commands between the server and the modem.
>>
>> But it still can never be 100% that way because tcp/ip just plain does not
>> promise that anything ever will happen either instantly nor even within
>> any specific time, whereas the fax protocol does require and promise that
>> various things happen within various times.
>>
>> --
>> Brian K. White brian@aljex.com http://www.myspace.com/KEYofR
>> +++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++.
>> filePro BBx Linux SCO FreeBSD #callahans Satriani Filk!
>>
>> I did it!!
>>
>> I compiled this code :
>> ftp://ftp.lantronix.com/priv/svr_tty/svr_tty-README.txt
>>
>> created a config file,
>>
>> runned the daemon
>>
>> and fax is working!!!!!
>>
>> the only thing is that now ssh can't create ptyp ports or some strange
>> thing
>> that I thing is related to this program that locks a device.
>>
>>
>
> well , what can I say, you were right from the beggining, but I tried, Fax
> is sent with no problems, but the other side gets a fax that it's fine on
> some rows and them fades on other rows and so on :P
>
> So I'll keep tweaking it or getting one of those class 2 modems.
>
> thanks.

oi ...headsmack

Why are you bothering to use tcp/ip-serial for a fax modem when you know that it can never be 100% because of fundamental timing issues that can't be fixed or gotten around? It's a _bad_ system. Do not use it. The fact that it sort of more-ore-less works most of the time is actually part of it's badness. It makes you think it's working and so you are tempted to use it, and thus will always have some amount of problems, perhaps infrequent but that will never go away no matter how much futzing you do over the years. Call up vsi-fax or ifax and ask the experts for yourself if you don't understand on your own from the description of the issue as I've tried to explain it. Faxing is not like other uses of serial lines. What's fine for dialup/login/ppp/uucp/etc... is not necessarily good enough for faxing. I told you about the virtual device driver in the interest of full knowledge and understanding, not so you could latch on to that one fact and ignore all the rest and charge ahead and "do it wrong".

There is nothing more frustrating than trying to help someone attain a solid system while they are determined to do things that are known not to work or at least known impossible to be 100% solid. IE: Not many people will "help you do it wrong" at least, not the ones whos help has any value.

It'll work most of the time though so if that's ok with you then have fun, but don't expect much help debugging or improving performance.

--
Brian K. White brian@aljex.com http://www.myspace.com/KEYofR
+++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++.
filePro BBx Linux SCO FreeBSD #callahans Satriani Filk!