In the last ISP News Report, I provided excerpts from a recent conversation with venture capitalist Gary Morgenthaler about why he invests in network hardware and software companies whose products simplify life for network managers. Here are more snippets from our conversation about his views on service provider start-ups and the outlook for Internet innovation.Q. Can you explain your interest in Netli, a service provider offering high-speed Web applications delivery?A.\u00a0 We like to go back and re-examine the assumptions that have been built into the infrastructure to see if some of them may no longer apply. With Netli, there are assumptions built into TCP. Everyone uses TCP. It works. It scales. It's reliable. So the assumption is that it must be good.But the problem with TCP is that it's a protocol that's well over 20 years old and was designed to work in a different environment, where communication links were low-speed dial-up with uncertain quality and not reliable.\u00a0 So TCP has a variety of features that reflect that. For example, TCP has a slow start because it guarantees that packets can go through every time without error. But you don't need that slow start with every new packet. The latency induced by TCP causes significant delays. If you're in China, you're looking at 10 seconds or more.What Netli solves is that problem of subsecond response times by taking a fundamental re-think of the underlying premise of TCP. When companies redirect traffic to Netli, they get an order of magnitude performance improvement. This creates the ability for corporations to provide [Internet] applications to employees and to customers that are highly interactive.Q. Netli is a service provider, vs. a hardware or software vendor. In general, how interested are you in investing in service providers right now?A.\u00a0 The service provider market tends to be commoditized and tends to be very capital intensive. That's a bad prescription for a VC. Netli is the product of several Stanford PhDs and some arcane mathematics that bring to its network global, sub-second Web response times. And it does so with very lightweight infrastructure. The network build-out has been done - they have a global network and access points in six continents - and they did so with surprising capital efficiency. The service provider model is a great model, but only if you can minimize the capital infrastructure and if you have a defensible innovation. We ultimately concluded that Netli has both.Q. I didn't see any investments in Morgenthaler Ventures' portfolio in wireless or VoIP. Why aren't those areas of interest to you?A. Part of the art of our business is figuring out when markets are poised for growth and making investments just prior. When markets are known or fall under the category of already being crowded [we tend to avoid them.] Wireless and VoIP are categories that are hyper competitive.Q. Explain your philosophy of investing. It seems like your trying to find more off-beat segments of the network industry.A.\u00a0 When we're doing our job right, we're seeing the world differently from other people. And we're seeing it the way it should be - a way that's simpler, more elegant and more efficient. We are funding entrepreneurs and assisting them in building industry-leading companies with innovations that make the world simpler, more efficient and more economical.Q.\u00a0 Because of the network industry downturn and the lack of investment dollars going into early-stage network start-ups, some observers say that innovation in the network industry is dead. What is your view?A.\u00a0 I think that's ludicrous. Innovation is anything but dead. Frankly, we can hardly keep up with the stream of really talented entrepreneurs wanting to see us and seek capital from our firm.I think the fundamental source of innovation is going to come from putting more and more electronics on smaller chips. One company that I should mention in our portfolio is Cortina Systems, which is an integration-on-silicon play. They are taking all the packet processing and frame processing as well as a transceiver interface...so that you have the entire Layer 2 on a chip. That used to be done on a board, and it was an expensive, complex board. Now it is on a single chip. Cortina's chips have found their way to market in Cisco's new carrier-class core router. We expect that technology to be standard, both within the core of the Internet and within the enterprise. We were the lead investor in that company.