Thursday, August 16, 2012

Found while following Apple/Samsung antics

A long, long time ago, I put my feet up and imagined what could happen given the trajectory of several technologies.  I'm not claiming anything, but I continue to be chagrined about the dysfunctions of the US patent system and the definitions of "novelty", "obviousness", "prior art", and "skilled practitioner".

This was published in 1998 (click near the page number for larger view):

Now, back to my popcorn.

Tuesday, June 19, 2012

Scratching the Surface

It requires a certain critical mass for me to emit a blog post.  Yesterday's unveiling of the Microsoft Surface devices has prompted copious product commentary, but I think the big story has little to do with speeds and feeds, or competition with Apple.

The Surface announcement is the best demonstration of the systemic weakness of the Wintel ecosystem, ever.  While multiple sources of pressure have been building to strain this long-standing duopoly, Microsoft's decision to manufacture its own tablet device may well push the water over the edge of the dam.

In addition, as a Microsoft shareholder, I'm concerned about the impact of this announcement on the company's overall performance.  More on that later.

First, the industry impact.  If we set the wayback machine far enough, we can go back to a time when personal computer manufacturers actually added substantial value beyond the semiconductor components supplied by Intel and its competitors.

A company called Compaq, for instance, distinguished itself in the business market by spending significantly on R&D to make Intel-based hardware reliable and serviceable enough to be taken seriously by enterprise customers.  They were able to charge a premium for their work.

Compaq's competitors, like Dell, were happier to take Intel reference designs, which were not quite as robust, bend some metal around them, and compete on price.  Intel and Dell responded by labeling Compaq solutions "proprietary" and simultaneously adding features into their reference designs, reducing their value to commodity levels.  Dell and its ilk won.

Intel was also working hard to migrate more and more value of a PC into its chips.  They were fairly successful at this, ultimately capturing most of the value in a PC and leaving table scraps of margin for their OEM partners.  Years ago I imagined that Intel might just go ahead and put the OEMs out of their misery, but it dawned on me that Intel was brilliant in outsourcing the misery - why would they want it?

The other winner at the head of the PC food chain was, of course, Microsoft.  They basically collected a tax on every Intel-based unit sold by every company except Apple.  Life was good in Redmond.

Meanwhile, at the end of the value chain. PC companies were fighting over increasingly smaller scraps.   The shakeout killed Micron, Gateway, and, ultimately, Compaq.  IBM had the good sense to see the handwriting on the wall and rebuild its company around services.  Now HP and Dell are trying to catch up.

Meanwhile, Wintel-based product design has suffered from a combination of consumer malaise and OEM fear, resulting in underspending in R&D.  Above the chip level, Intel has maintained a leisurely  R&D pace, with the ultrabook reference design a recent example.  But, as we all know, Apple spoiled the party by creating a new category and nailing it with the iPad and its app ecosystem.  Android tablets piled on.  Wintel OEM response has been ham-fisted at best, partially because neither Intel nor Microsoft was ready with a fast response.

But Wintel OEMs are also fighting for their survival, are being pulled in multiple directions from external and internal forces, and are shedding resources like clothing on a hot day.  Microsoft must have felt that, left to their own devices, it would take so long for OEM partners to come around that it would be impossible to play catchup.

In addition, Microsoft could do the same bill of material arithmetic everyone else was doing and conclude that, for some devices, ARM had a place and an Intel premium was not worth paying.

The result - Microsoft takes matters into its own hands.  It uses an ARM processor to compete on price, and an Intel processor to ride its Office monopoly.  These are not dumb moves.

But,  Microsoft has hit its OEM partners squarely in the face.  This is not a Zune. nor is it a Microsoft mouse, each of which could have complemented an OEM product.  If successful, Surface devices will take lots of dollars away from mainstream OEM products.

Intel probably views the ARM-based Surface as a minor annoyance.  They have seen Microsoft move to other processor architectures before (NT on Alpha, anyone?), and know they can usually play the long game and win business back.

But the PC OEMs are a different story.  I'd hate to be at HP while I watch the last vestiges of any proprietary advantage in tablets with WebOS walk out the door.  I'd like to think that Google might use this event as a reason to really woo PC OEMs with a non-balkanized version of Android, but I'm not holding my breath.  Besides, Google would have to come up with a unified Android app marketplace that OEMs could white label, and share revenue.  That could happen.  And pigs could also fly.

It's no sure thing that Microsoft will make Surface work.  But let's suppose they do.  As a MSFT shareholder, I can therefore look forward to some increased revenue, and lower gross margins as hardware gross margins become a more significant part of the Microsoft product mix.  And I'm wondering, why in heck isn't there a new arms-length company being formed to do this, so margins aren't diluted across the enterprise?  Intel continues to outsource value chain misery, and here is Microsoft jumping to take it on.

For an outside observer, the new array of moving pieces created by this announcement is awe-inspiring in its complexity and unpredictability.  The Surface announcement party may be over, but the real fun in Wintel-land is just beginning.

Tuesday, April 17, 2012

Rational Economic Behavior and the Internet: Why You Want to Pay Per Packet

Originally published in 2010.
Verizon has raised my rates since then.  Grrr.
This bolsters my argument.

It’s always dangerous to give me spare time, an Internet connection, and a calculator.  I have been thinking about this particular issue for many years — any kind of “all you can eat” model has always struck me as suboptimal, and my reaction to the growing popularity of flat rate Internet connection pricing based on connection speed is the same as my reaction when offered free shipping on 50-pound bags of dog food back in the day: buy all you can, ‘cause this deal can’t last.

Internet customers (at least in the US) seem to have figured this out.  Their usage of bandwidth continues to grow at very high rates while flat-rate pricing is the norm. This is rational producer and consumer
behavior; there is neither disincentive to produce “fat” content or applications nor disincentive to consume them. This has been a lurking but invisible problem in the wired world, where bandwidth is relatively cheap. But the problem has become visible recently in the wireless world, where bandwidth is more expensive (because it is more scarce). We have all seen reports of wireless carriers struggling to keep up with network demand created by the latest smartphone or media application.

It costs real money to upgrade networks to keep pace with this demand, and those costs are ultimately borne by the subscriber. So in the US, we have carriers trying to raise their rates to offset increases in capital and operating expenses to the point where consumers are beginning to push back, and the shoving has come to the attention of the Federal Communications Commission, which has raised the possibility of treating Internet network providers as common communications carriers subject to regulation.

I believe that flat-rate pricing is a major source of problems for network carriers and consumers. In the carrier world, the economics are known but ignored because marketers believe that flat rates are the only plans consumers will accept. But in the consumer world, flat rates are rising to incomprehensible levels for indecipherable reasons, with little recourse except disconnection. Consumer dissatisfaction is rising, in part because consumers feel they have no control over the price they have to pay. This is driven by their sense of pricing inequity that is hard to visualize but comes from implicit subsidies in the current environment. The irony is that pay-per-use pricing solves the problem for carriers and consumers.

The Current State
Let’s do some arithmetic. I went to the statistics page on my router, where it reported that in the past 47 days I had sent and received around 15 million Internet packets. I rounded that up to 20 million,
then derived the following:

         - 20 million packets in 47 days
         - 425,532 packets/day
         - 12,765,597 packets/month

How big is a packet? Like many questions, the answer is, “It depends.”  After a little research, I determined that my average packet size was somewhere between 557 and 1,500 (an average rich-media packet size) bytes. Then I made some assumptions about my traffic:

         557 bytes (60% of traffic)
         - 1,500 bytes (40% of traffic)
         -  934 bytes (weighted average packet size)

Now I had an average packet size, and I had my packet traffic per month. So how many bits is that?

         - 7,474 bytes (weighted average bits/packet)
         - 12,765,957 packets/month
         -  95,412,765,957 bits/month

Whoa, that’s a lot of bits. Let’s scale it down to manageable significant digits:

         - 90,993 Mb/month
         - 11,374 MB/month
         - 11 GB/month

This exercise revealed that, in a typical month, I shipped about 11 GB over the Internet (the vast majority was downloaded but for this discussion we will focus on total traffic). Verizon, my provider, charges me about US $64.99 per month, or about $2.17 per day for 25 Mb/sec download speed.2 Some of you might say, “Hey, $2.17 per day is cheap.”  My response is “Compared to what”? Let’s take a look on what I actually consumed:

         - $64.99/month
         -  $2.17/day
         -  $5.85/GB transferred (actual)

Is $5.85 per GB a good or bad deal? Let’s compare that price based on actual usage to the theoretical price I could be paying if I used all of my monthly bandwidth:

         - 26,214,400 bytes at 25Mb/sec peak rate
         - 3,276,800 bytes/sec
         - 11,796,480,000 bytes/hour
         - 11,250 MB/hour
         - 11 GB/hour
         -  0.0107 TB/hour
         -  0.2575 TB/day
         - 8 TB/month

Now, all the carriers disclaim their peak performance numbers; they can be degraded by traffic congestion, phases of the moon, whatever. Let’s pretend that I could actually see the above (and, to be fair, my FiOS service has been generally reliable). That would mean I could transfer 8 TB in a month, compared with the 11 GB that I actually transferred.  That changes the transfer price dramatically:

         - $64.99/month
         -  $5.85/GB transferred (actual)
         - $0.008/GB transferred (theoretical)

Clearly, I should be consuming more. Even rounding up, the price gap between $5.85 per GB and $0.008 per GB is substantial. And herein lies the problem. Since the rate I pay is flat, everything in the gap between 11 GB and 8 TB looks free to me. But I assure you it does not look free to Verizon. If every FiOS subscriber tried to download 8 TB a month, cries of panic would echo throughout the land. A similar exercise for wireless data plans would show higher prices but a smaller gap between actual and theoretical because the theoretical limits are lower.  Some wireless carriers explicitly prohibit extended peak bandwidth use for just this reason.

A Monthly Bill I’d Like to Get
What if, instead of increasing the maximum download speeds for a flat rate, a carrier tried this: pay $0.0000025 per packet — period. The resulting bill would be:

          - $31.91 for 11 GB transferred/month
          - $63.82 for 22 GB transferred/month
          - $22,979 for 8 TB transferred/month

A pricing strategy like this might lower the monthly bill for an average customer as well as allow for some increased consumption without breaking the bank. It would provide the consumer a way to lower costs just by reducing consumption. There would be a substantial disincentive for order-of-magnitude increases in consumption; if the data were really that valuable, this would prove it. Otherwise, consumers would have to think of options:

         1. Do without.
         2. Find a way to make the data more efficient.
         3. Find a cheaper substitute.
         4. Get someone else to pay.

All these options are used by producers and consumers today and form the basis of advertiser- and business-supported messaging. There is no reason to suspect that a motivated advertiser wouldn’t pay to transmit its car ad to a serious prospect. Web site and content creators would be obliged to consider data transmission costs and build more efficient products — or offset their higher costs with direct payments, advertising, or some other explicit subsidy. Carriers could use variable pricing per packet to adapt consumer usage to network capacity variances and expansion.

This past June, AT&T, wireless provider for Apple’s iPhone, became the first major mobile phone company to stop offering new smartphone customers a single monthly price for unlimited Internet access.3 That may signal an industry shift to charges based on how much people use their phones to access videos, music, and data. AT&T expects the new pricing to boost sales. Verizon followed in July.  Newcomers to AT&T can pay $15 per month for 200 MB of data or 2 GB for $25. AT&T says 65% of its smartphone customers use fewer than 200 MB per month, and 98% use fewer than 2 GB. Just 3% of AT&T’s smartphone customers account for as much as 40% of its data traffic. With the limited airwave spectrum available for wireless broadband, other providers may switch to usage-based pricing, including Sprint and T-Mobile. Other businesses may be getting the message as well. Silicon Alley Insider recently reported on the rumored new Apple iTV with the headline, “Apple's New iTV Could Finally Force ISPs to Give Up on All-You-Can-Eat Internet Access and Jack Up Your Bandwidth Bill.”

In short, rational economic behavior would prevail in a pricing environment free of implicit subsidies. This would prove beneficial to content producers, content consumers, and the infrastructure providers that move the bits around. When I look at the problems created and looming by current flat-rate pricing — and consider the advantages of usage pricing — I believe it’s only a matter of time before usage pricing becomes the standard.

In the meantime, use all the bandwidth and transfer all the bits you possibly can. 
This deal is too good to last.

1 - Slaptijack. “Average IP Packet Size.”  
      Facebook Note, 18 March 2010 (
2 - Verizon offers Web mail and other services for this price. For simplicity, I have ignored all bundled 
      services in this example.
3 -3  Lieberman, David. “New AT&T Smartphone Users Won’t Get One-Price Net.” USA Today
     4 June 2010.
4 -Frommer, Dan. “Apple’s New iTV Could Finally Force ISPs to Give Up on All-You-Can-Eat
     Internet Access and Jack Up Your Bandwidth Bill.” BusinessInsider/Silicon Alley Insider, 
     23 August 2010 (