As a keynote at yesterday’s VoIP Developer Conference in South San Francisco, the fresh-faced 27-year-old Spencer, wearing khakis and an open-collar shirt, looked the archetypal entrepreneur who is as common in Silicon Valley as, well, silicon.

But his open-source Asterisk project, which began in 1999 when Spencer built a phone system for his Linux-support employer using Linux, has legs.

Named after the wild-card of Unix and Linux, Asterisk is being developed by more than 300 coders and has nearly 250,000 users. This year, IBM and AT&T showcased Asterisk at some industry gatherings and Lucent currently is using Asterisk in its testing labs.

Essentially, open architecture such as Asterisk promises reduced network operating costs and complexity for VoIP and other IP-based applications, including e-mail and videophone. Open-source software, however, may potentially be more difficult for an enterprises’ IT staff to support.

Still, Open source enables resellers and integrators to customize existing hardware and systems, which adds value and, ultimately, margins, Spencer said.

Asterisk leads the small yet growing pack of open-source telephony projects, but that may well change during the next couple of years.

Asterisk is an open source, hybrid TDM and packet voice PBX and IVR platform with ACD functionality. Basically, Asterisk is a black box, a piece of middleware that connects any kind of interface to any kind of voice network, Spencer explained.

Asterisk supports encryption and authentication from RSA, he said.

But the biggest roadblock for Asterisk is without doubt quality, said David Mandelstam, president of Sangoma Technologies Corp based in Ontario, Canada, which makes PCI cards for wide area networks and TDM voice and video gateways.

[Asterisk] is not close to carrier grade, Mandelstam said. It’s free and we’ll live with its quirks because the alternative is just so much more expensive.

Asterisk also has not been designed for scalability, he said. Yet unlike its open-source rivals, Asterisk does actually work, he said.

Still, Mandelstam calls Asterisk the 9,000-pound gorilla of open-source telephony. They are leading by a long shot. Not just because it’s well marketed but because it’s the most complete and it works, he said.

Still, Asterisk would be entirely rewritten from the ground up during the next two years, he predicts.

Unlike Asterisk in its current form, which was written in a hurry to solve specific, isolated problems, according to Mandelstam, what would emerge would have comparable feature sets but would be expandable and of a higher quality.

This carrier-grade version may emerge as Asterisk or it may come from another open-source project, he said.

I know a number of people who are trying to do it, he said. Among them is the Aefirion project, which is working through and fixing Asterisk module by module. OPAL also is a project seeking to re-write Asterisk as being carrier-grade, but take a hybrid approach, using a combination of Asterisk and OPAL code, until the Asterisk code is completely rewritten.

Asterisk, however, has the benefit of Spencer’s small company, Digium Inc, based in Huntsville, Alabama, which sells commercial licenses of Asterisk for enterprises that don’t want an open-source license, as well PCI cards.

The 37-employee company sells its wares through more than 130 Asterisk systems vendors.

Digium has been profitable since 2002, Spencer said. Boutique and heavyweight venture capitalists have approached Digium, but he has not yet decided on outside funding.

The privately held company said that between 5% and 10% of all Asterisk installations use Digium hardware, but declined revenue and other financial metrics.

Just over a month ago, Digium released its Business Edition version of Asterisk, which sells for $995. The company also has a subscription model for an annual fee.