"Linux Gazette...making Linux just a little more fun!"

Those of us that have used Netscape (or other Web browsers) under Windows, may have felt some envy at the sight of the Dialer, a little box in one corner of the screen showing that you are on-line and how much time you have already spent on-line, so your phone bill doesn't go overboard.

In Linux, on the other hand, setting up a dial-up connection and making it work is often a rather painful process, a "challenge", if you like: Not only no handy auto-install packages available from your internet service provider -- you have to figure out everything for yourself, and know what questions to ask -- but also establishing the connection every time requires you to go through a sequence of operations.

Open an xterm or a virtual console, log in as root, and run the ppp startup script (unless of course you use the  diald package for dial-on-demand, an alterative also. I personally found that it had too much a mind of its own :-).

Closing the connection similarly requires you to do the same to run a disconnect script.

One of the first things I did therefore when I decided to learn tcl/tk was to write a Dialer look-alike. It (tkdial) is attached to this text; it is the first tcl/tk program I ever wrote -- just under 150 lines -- and that may show.  But tcl/tk is ideal for this kind of job, "glueing" existing command line facilities together into a beautiful motif-look, mouseable package. Just have a look at the pictures!

 

There are some things with a dial-up connection which appear not generally known (I'm  not talking to you, geeks and gurus :-). I'll give a quick run-down of my experiences as I understood them (but note that I am no professional):

• In order to be able to run tkdial (which calls pppd) as an ordinary user, you should have pppd set suid root. Additionally, you should be able to read the scripts in the /etc/ppp directory, so they should either be world readable or readable by a group to which you belong. (A nice exercise in basic system administration. But if you give world reading rights to your pap-secrets file, you will deservedly fail your exam!)
• The standard Red Hat sendmail setup uses sendmail -bd -q1h, in other words, activate the sendmail daemon once an hour. That's not much. In a dial-up environment you want to send out mail while the line is up, so change the -q1h to -q2m, for example, for every two minutes. And follow with the mailq command if your mail really has left your machine, before closing down the ppp link. (If you forget, not to worry: The queue will continue to try for five days, every time ppp comes up.)
• There is an option to pppd called lcp-echo-interval, which can be used to keep the line alive. LCP means Link Control Protocol, and by putting an option lcp-echo-interval 60 into either your /etc/ppp/options file or on the pppd command line when starting it up (i.e. inside the tkdial script file), you can keep your line alive even when not actively browsing.

This is important because, with the ubiquity of crashy operating systems, internet service providers have taken to the habit to cut the line when nothing has arrived over it for a couple minutes. Imagine starting a five hour download, going shopping, and returning only to find that three minutes after you closed the door, the machine crashed and the phone line is still open, burning up your money for nothing! (This could even happen in principle with Linux, if the power goes down and you don't have an UPS, or your dog gnaws off the phone wire. Well, the modem has a time-out also). So Windows dialers send an empty package once every minute or so to the ISP, telling "don't worry, I'm still alive!" And when the system crashes, the line cuts promptly.
With the option given above, also Linux will send an empty package every 60 seconds.
• If you have a POP3 mail service, the best program (transport agent) undoubtedly is fetchmail, which transports the mail to your "system maildrop", typically /var/spool/mail/<userid>. Also fetchmail can be run as a daemon. You can use xbiff or xmailbox to inform you of arrived mail, and read it with pine, exmh (recommended, another one of those tcl/tk miracle programs!) or whatever. If you use Netscape mail, forget about all this, you just have to configure it on its own terms, which involves learning pretty much the same concepts anyway.
• A trick (I don't really know if this is wise or intended, but it sure is effective!):

If you use very much the same search agent all the time, e.g. Alta Vista, put it in the file /etc/hosts. Find out Alta Vista's IP address with ping. The details are left as an exercise for the reader, as well as the explanation for the speedup achieved (hint: DNS...)
• Make sure that your machine name (as given in the network setup procedure) is the same as that which your ISP gave to your mailbox. So, if you are [email protected], call your machine isp-international.com. Not very romantic, but you avoid problems with an anti-spam feature in some sendmail installations, which bounces mail coming from a "sender" not existing (i.e. not found by the domain name service) on the internet. (I expect that this feature can be circumvented by reconfiguring and recompiling sendmail.cf. I guess the sendmail folks just bet that such a feat is way beyond your average spammer, and I bet they're right...)

Alternatively, make yourself exist; but that requires the co-operation of your ISP. E.g. EUnet would give you a mailbox name of [email protected], which provides you with a slightly more personalized name for your own machine...
And make sure to keep the localhost  name also valid. Some programs depend on it.

Acknowledgement: I am indebted to Jaakko Hyvätti of EUnet Finland, who provided me with working ppp scripts and plenty good advice.