Response to Jonathan’s “Sun’s Network Innovations (3 of 4)” article

I posted the following comment in response to Jonathan Schwartz’s latest blog post and vlog “Sun’s Network Innovations (3 of 4)“, however Jonathan and his blog pixies can’t be up and reviewing this yet because it’s awaiting moderation (and too right too, it’s the dead of night in California, and frankly I’m happy Jonathan is keeping busy running Sun, as he’s our CEO, if you didn’t know).

Hi Jonathan,

In 1998 I was involved in building Harrods Online as Chief Architect (employed by Harrods). We used a heavily Netscape based stack, using both Netscape Web Server and Netscape Application Server. And at the time Netscape had over 56% of the Enterprise Web Server market-share as well as being some of the best available technologies at hand.

Both of these technologies were available under free to use licenses from Netscape, with payment for support on top. Well I have to tell you in three years running this stack in production we never once paid for support, not a bean, nada, not on Christmas day, not on any day. Why? Because these were such stable technologies performing in their core capabilities that we knew from direct experience that very little would go wrong and because we architected a fully redundant and highly-available architecture based around horizontal scaling, network load-balancing, and avoiding session persistence where possible.

Looking back I would have liked to have paid Netscape something, even a token amount, because they were a great company, sadly missed, and perhaps if they’d been able to generate more revenues they’d still be with us. However this was a completely economic based business decision so we didn’t. Of course Sun eventually acquired these technologies through the iPlanet Alliance and we were able to integrate some of the innovation, robustness and domain knowledge into our own technologies.

I really wish at the time MySQL had been available, because we wouldn’t have had to pay Oracle either (our biggest license cost outside of Vignette, which was a core component). And before you ask, no, Harrods Online never went down because of any of the Netscape components, if anything by far the biggest problem areas were Vignette (esoteric, overtly Vignette specialised skill set required) and Oracle (connectivity to Vignette, Netscape, the other technologies we used, a number of content version issues, and it’s applicability as a web-scale technology). I know that some of these issues in these technologies have been resolved, but in both cases this was true then.

What’s cracking is that Sun has some of the best Architecture Services for Open Source deployments, especially large-scale implementations. Sun have a genuine sweet spot in the three areas needed to get these to work in the most cost efficient manner: Open Source Software (low barriers to entry, low barriers to exit, ease of availability, applicability to function), cost effective and highly price performant Hardware (Servers and Storage, both Open Source and Open Standards based), and the Professional and Managed Services needed to tie it all together and ensure operational effectiveness.

Frankly, I suppose we were ahead of the pack in many ways at Harrods Online, because the architectural principles used there are precursors of the one’s you’d see in a modern web-tier architectures, such as Flickr or Delicious. I know that if I was building that now not only would we have a near on 100% Open Source stack, but I’d also be looking at how I could offload some of our costs (especially in terms of low level and non-functionally focused skills and technologies) by utilising a Cloud Computing based infrastructure and platform provider.

All the best,


If, and when, it gets authorised I’ll add a link here.