Fax is everywhere because the "hard dollar" costs are so low. For less than $100 you can get a fax machine, plug it into a standard phone line and you're on the network. Faxing requires practically no training, and it's like learning to ride that bicycle: Once you've learned how to use one machine, you can use them all.

More important, the fax sender pays the cost of delivery; it costs nothing to receive, so as a user you're in complete control of your fax bill. With an economic incentive to be sure the receiver will be interested in the message, "fax spam" remains at relatively low volumes, posing nowhere near the same problem as unwanted e-mail. And yet too much of what we receive via fax is useless--even if we're interested in the information.

The problem is that "soft dollar" costs to the fax receiver are high. Every day I receive a stack of faxes, which I must sort through by hand. I immediately discard a third of the paper (the cover pages) and then plow through each page. I get a strong urge to throw all this paper away because no filing system known to man can do it all justice.

A network-based fax server can help the sender significantly. It's easier to submit and manage large jobs, control delivery costs and monitor usage with a centralized, outbound fax server. Distribution list management is a breeze. E-mail and mail-merge integration is mature. Preparing a transmission for 100 recipients takes little more effort than preparing a cozy little transmission for two. In fact, a commercial fax service can deliver a message to hundreds of recipients simultaneously.

Spinning Wheels But on the receiving side, fax servers are a failure. There's simply no foolproof, cost-effective way to automate fax distribution over a data network. Vendors and standards bodies have proposed several solutions, such as subaddressing schemes: Why not ask the sender to enter a few extra routing digits? Because the PSTN doesn't buffer the tones at dial time, forcing the sender to wait for a good connection before entering secondary routing codes. And it's difficult to change senders' usage habits. How about OCR? It's not reliable for printed text, let alone handwriting. Maybe DID (Direct Inward Dial), which assigns a virtual phone number to each user? Sorry, it's expensive and doesn't scale (yet another missed opportunity by telecommunications carriers--DID is not well-understood and is priced beyond what users are willing to pay). The last and only reliable option is for manual routing by a fax administrator--a ridiculous waste of time and effort that slows delivery.

Even if we get past the routing problem, electronic fax formats are useless and bloated, and for that we have our standards bodies to thank. The ITU-T's recently adopted T.37 format (coinciding with the IETF RFCs 2301 to 2306) describes integration between e-mail and fax, using standard RFC 822/1123-style addressing. The IETF's defined MIME type for fax attachments to SMTP messages is a TIFF (Tagged Image File Format) image. Unlike, say, Adobe's PDF, TIFF has no concept of text mixed into the image. To make matters worse, the TIFF version required by the IETF produces a raster image that is suboptimal for data networks.

A typical TIFF attachment using common Modified Huffman coding (the base requirement in the RFC) consumes an average of 100 KB per page, and I can't feasibly edit the darn thing. Since the computer doesn't recognize any text here, I can't index the file's contents. I can heave this beast into a directory somewhere, but without indexing, who knows how I'll ever find it again. So while fax to e-mail makes the message portable, it doesn't really help me manage information.

Accepting these limitations as they are, perhaps I should simply forget e-mail integration. By converting incoming faxes into Web-based documents, I can access them from anyplace I can launch a Web browser, which is to say just about anywhere. Products such as Optus Software's FACSys Web Agent and RightFAX's Web Client let me view and manage faxes this way, keeping the clutter out of my e-mail-- solving the portability problem without incurring the bloat from TIFF files.

Riding a Schwinn Down I-95 On the transmission side, numerous devices deliver fax over data networks--for ATM and frame relay networks especially. These products can reduce the cost of sending faxes, but are confined to use within private intranets, where e-mail is preferred over fax. Too bad, because this is a technological feat. While you might think delivering fax over a data network is trivial, technically it's not. The T.30 protocol used by Group 3 fax machines is quite timing-sensitive, and fax-over-data devices must perform complex operations on the protocol--modulating and demodulating the fax traffic at endpoints or spoofing the protocol for the remote fax machine. Either way, this complexity makes it difficult to build fax support into devices inexpensively.

A handful of proprietary IP-based fax devices already exist. First, there are gateways that eliminate the need to replace existing PSTN-based fax equipment. Most fax machines aren't physically modular or software-upgradable either, so an external gateway is required to do this trick. These devices speak T.30 to the fax machine, buffering data to be sent over the data network when the remote destination is IP-ready.

The Internet Magic Internet Fax Legacy 2000/3000 fax gateway products sit between a standard fax machine and an ISP or public switched network connection. If I can reach a remote fax machine equipped with another Legacy device, I send it for "free." Using T.37, it can deliver faxes to e-mail accounts, or receive faxes sent by e-mail. And at $499 (or $599 for a LAN-based gateway), it's about half the price of its closest competitor, the @fax FAXfree FAXportal 500. That's not too expensive, but still not sufficiently low-priced for mass appeal.

The first all-in-one IP fax machine is Panasonic's Panafax UF-770i. It's a full-featured fax machine that connects directly to an Ethernet network, delivering faxes to other UF-770s, e-mail accounts via T.37, or standard fax machines attached to the PSTN or an @fax device. When we brought this device into our labs we were excited: "An Ethernet-connected fax machine?! Woo hoo!"

Then we tried to get it going. We've got guys here who code in Java but who couldn't figure out the LCD interface on this thing. Or more accurately, couldn't be bothered--after all, this is supposed to be a fax machine! The unit violates the two fundamental laws of fax--it must be simple to use and inexpensive; Panasonic's Panafax is neither.

In the interest of making real-time IP fax delivery more widespread, the ITU-T has introduced the T.38 protocol, specifying an open standard for delivery of Group 3 fax messages using either TCP or UDP, at the vendor's option. Using UDP may simplify the protocol stack requirements compared with using TCP, but using UDP will likely increase the complexity of higher-layer software required when packets arrive out of order or are lost.

T.38 has been adopted as the standard fax coder for H.323, in much the same way that H.323 voice over IP products select from various audio codecs. Vendors such as Dialogic have already demonstrated versions of VoIP gateways with T.38 fax capabilities.

The pricing of T.30 to T.38 gateways--or the services that deploy them--must be very low or the technology will never go anywhere. Internet Magic, for one, is promising T.38 in its products via a simple, free software upgrade. So $500 is the price ceiling for now, but prices must drop further--much further--for adoption into the mainstream.

At the Finish Line What effect will open standards for fax-over-IP produce? Once again the sender will gain the most. In private networks, standards-based products could participate in the least-cost routing schemes offered by traditional fax server vendors. This will let smaller offices join internal fax distribution networks at much lower cost, using products like Internet Magic's.

But remember the economics, and consider how this upsets the balance. T.38 promises to reduce the cost of sending a fax to practically nothing. T.37 will allow e-mail users to easily address fax machines, and fax machines to fill e-mail boxes with images. People who don't even have e-mail accounts will generate e-mail spam from a fax machine. More clutter, more bloat and less manageability--does this sound like an improvement? Perhaps things are better just the way they are.

Send your comments on this column to David Willis at dwillis@nwc.com.