(this is page 95 of 121)





Latest

Jon Udell has a great interview on his podcast "Interviews with Innovators" where he talks to Ed Iacobucci, a veteran software guy (founder of Citrix) about his new business, DayJet. Like others before him, Ed is looking at the constant money-loser, customer-displeaser that air travel has been in the United States and asking the simple question: does it have to be this way?

But unlike most others, Ed came out of retirement to start a business around demand aggregation for seats on a network of much cheaper regional jets (made by another software industry veteran Vern Rayburn). The core idea is that the network provides the ability for customers to self-select along a "time arbitrage" spectrum, and that when this is combined with some heavy duty optimization algorithms, the capacity of the overall system can be managed to make what has thus far been a terrible business (regional charters) profitable.

I won't go into the specifics of how this all works because Jon does such a good job of covering it in the podcast (I'm really digging Interviews with Innovators by the way). What I will say is that it is really interesting to see people that come out of the software industry try their hands at other businesses while borrowing a lot of the software thinking that helped them before. Outside of helping him to see two core business problems from the computational perspective (a potentially huge advantage), I was left wondering how much Ed's years at IBM and then Citrix are influencing his thinking in this new endeavor.

I first met Ed during the time when he was having his "false start" in regional air travel. He was brought to our office at Memora by a common friend to see if what he saw us doing would make for an interesting investment. And even back then, my take-away from our meeting was that he was very interested in how we might make a run at the platform that Microsoft was putting together for attacking the home. Though he was a very nice guy, he was rather skeptical about our prospects for making any kind of significant dent in the market. In the end he was both right and wrong: right because we couldn't make it work (Memora), but wrong because worrying about the Microsoft platform (or any platform) proved to be entirely useless. Six years later, the digital home is still unconquered, and the company that looks closest to it today had nothing but an overpriced portable music player which people were convinced was named after the acronym "Idiots Price Our Devices" back in 2001.

Throughout the interview Ed spends a bit of time talking about platform he is trying to usher in— at first making the analogy between the airframe being like the PC, the engines being like the CPU, and DayJet being like the OS/application suite— which is an interesting analogy which I found lacking in its predictive power. Interestingly, I find a lot of the same stuff going on today as software veterans talk about the Internet as a platform, or the mobile handsets as platforms— interesting but not totally insightful.

Meanwhile though, I think I am beginning to develop an aversion to the term platform.

Comments Permalink

A few weeks ago while trying to find someone in the middle of a downpour in New York City, I borked my iPhone (iPhoners be forewarned, apparently, the smooth surface and lack of buttons has little to do with the iPhone's ability to resist water). Since I was away from home and in need of a phone that could make and receive calls, I went to the local Apple store and had it replaced with a loaner unit. Despite my expecting it, the most amazing part of the experience was what happened when I plugged the loaner into iTunes— most of my phone's state was restored from a backup including the state of the various messaging applications (SMSes, emails, read/unread, etc.). While I lost a few things (YouTube bookmarks and ringtones), the overall experience underscored the fact that more and more, the actual devices that we do our computing on are nothing more than interface and an intelligent cache, while the bulk of what matters to us today (data) is increasingly moving to the cloud.

Ok, so I cheated in that paragraph above a little bit. Most of my iPhone's state was actually stored on my Mac laptop, which today is not strictly in the cloud. In fact the Mac makes a bad case for the cloud, because its own state would be so difficult to replicate up to a server farm somewhere. Sure, I can use DotMac to backup files and the datastores of specific Apple applications (Addressbook, iCal), but most of what represents my Mac (my iTunes collection, my tweaked out DarwinPorts, my various .rc files, the rest of my installed applications) doesn't play well with cloud-based services because we currently lack a good protocol for replicating the data and machine state to the cloud.

Or do we?

I guess I must be having a bad month, because the main server I use for experiments (my own personal bit of the cloud) seems to have gotten sick somewhere in its bowels. Every few days the kernel panics and the entire machine locks up. After re-jiggered most of the software installed on it, I've now come to realize the root cause is some bit of flakey hardware on the motherboard or RAID controller. Having probably installed Linux more than 100 times over the last 10 years, I can honestly admit that the magic of building a useless bunch of PC parts into a little piece of the Internet is now all but gone, and that even in the cloud, it's probably a good idea to treat the hardware as ephemeral.

Enter virtualization. A few months ago, I remember seeing Brad Fitzpatrick enamored with the open source para-virtualization project Xen because he could just "spin off VMs for each application." However, at the time I didn't have the cycles to play with it. The intervening months have done little to make the open source part of the Xen project move forward (there is still no bootable CD version of it, most documentation is at least six months out of date, and XenSource is now on the way to the dustbin of acquire-to-kill companies). Since I didn't have quite enough patience for digging through all of that, and since I have been quite impressed by VMWare's Fusion (the Mac OSX version of VMWare's VM runner), I decide to give their free server product, VM Server, a shot.

[A parenthetical aside on those interested in Xen versus VMWare on a technical level— all others skip this paragraph: VMWare Server doesn't quite perform as well as a fully "paravirtualized" VM (like Xen) because it has one more level of abstraction between the guest OS and the hardware. In effect, a guest OS on VMWare Server runs as a process on a base (or host) OS without modification to the guest OS, and with minor tweaks to the host OS (implemented through a set of kernel modules). It's not like those old crappy CPU emulators (Virtual PC comes to mind here) in that instructions are still executed natively by the processor (which is why VMWare primarily supports Intel on Intel configurations), but you will pay a penalty, especially around IO operations. That said, for my workload, the penalty was neglible. In fact the server that pushed this page to you is running in a VMWare instance inside of an Ubuntu Linux based OS— go ahead, hit reload really fast a few times to see how much the network has become the bottleneck these days!]

To make a long story short, I now have a bunch of Linux VMs which I run on two physical machines, as well as my development laptop interchangeably. In and of itself this is nothing special, except that maybe it is interesting to see how some of us like to waste the cycles that Moore's Law gives us with an increasing number of levels of abstraction.

But back to getting my Mac into the cloud, for real. Why couldn't the OS support virtualization natively, and then let me run all of my computing on a guest OS that itself could be streamed up into the cloud? Wouldn't this then become the ultimate (albeit complicated) protocol for syncing state between client devices and the cloud?

There are interesting possibilities around a solution like this one that go well beyond backup and restore from the cloud. At a crude level, this VM-shipping is the same thing as packaging up code and data and moving it around the network. It is in effect an entire execution context that is just mine. Does it mean no more traveling with a laptop? Sure, but it also means being able to keep my computer running all of the time, whether I have it on or not. This isn't anything new for the energy-sucking desktop crowd, but as the world moves to laptops, there may be some interesting new businesses here.

To close, let me come full circle back to my borked iPhone. Naturally, there is little chance that a device that is as power and cycle-constrained as the iPhone is going to grow a virtualization layer any time soon (thought it's amazing what can be put on chips these days). The more relevant question though is: should it? Are there benefits to having every computing platform that we use be able to save its entire state in one bag of bytes that is portable, cloneable, and ultimately, sychronizable across other devices?

I'm not sure, but then again, I've definitely grown up in the simplest possible protocol age where the best thing you can do for an application is emit its data as an RSS feed. It will be interesting to think about what happens when these virtualized devices find more comfortable homes in the cloud.

Postface: I wrote most of this post at 2 am while recovering from my brother's 42nd birthday celebration but then lost it to a fibbing laptop battery (fittingly enough). It's his total and utter lack of respect for conventional wisdom (combined with more than a few drops of crazy) that I've found most inspiring in thinking more broadly than the immediate engineering problems in front of me, and for that reason this wacky entry is my way of saying: happy birthday brother.

Comments Permalink

Blackberry Internet Mail service seems to have borked this morning, leaving millions of Crackberry addicts without an excuse to avoid their children. I'm sure in the end it will have been some silly admin error that caused it, but as we grow more and more used to (and addicted) the gigantic signaling fabric that the Internet has become, it is going to be harder and harder to cope with these kind of outages. Why just last week, we were just getting over the Skype outage during which I heard podcasters claim they were going to be "put out of business" by bad Skype capacity planning.

What is especially rough about the Blackberry outage is that it is endemic of the types of outages we can expect to see if we continue switching to proprietary (and thus centralized) messaging platforms. For instance, all sorts of web services have sprung up to replace email with varying degrees of success. On some sites, the internal messaging is peripheral to the experience (Flickr), but on others, it is either central (Facebook), or the experience (Twitter). Unfortunately, these systems were designed as centralized web services. And even well implemented web services can go down, mostly because the redundancy is added after the initial design. For instance, I'm sure that Facebook's database layer is sharded and replicated, its assets are on multiple CDNs, and its application servers are geographically distributed— but because the application was never designed to be distributed, these solutions will only take robustness thus far (especially after you add in the potential for human error at each of these layers).

Contrast this to decades-old email, or rather email + DNS. The combination of a very simple protocol for message exchange (SMTP) with a very flexible address resolution system (MX records in DNS) means that is is pretty hard to make all of email "go down." Sure the big centralized services like Gmail and Hotmail can have outages, but all of those law firms running Exchange, schools running Sendmail, and startups running Postfix will keep right on chugging. And even when these small poorly managed mail servers fail, there is usually some marginally better server at the ISP queueing up email just for the occasion.

Ray Tomlinson and his colleagues at BBN deserve a ton of credit for having put together a system in the 1970s that could grow to the scale and scope that it has while remaining more robust than a lot of what has come after. All of the rest of us should take a page out of their book in working towards improving the fabric of the Internet.

Comments Permalink

Get social camera!

I find myself taking either of my real digital cameras out less and less these days which is a little ironic, given where I've just come from. Actually, I still use my Nikon D70 in much the same way (episodically around events where I want to get a set of very permanent pictures), but my small Canon point-and-shoot rarely sees the light of day. It still takes much better pictures than the iPhone, and over the last week I've found myself wanting the zoom and low light shooting a couple of times, but overall, the combined needs just don't get me over the activation threshold (which is telling given how small these little cameras have become).

The other thing I've noticed is that I'm just not super excited about the category of the small, disconnected point-and-shoot camera anymore. It used to be that every 6-9 months I'd find something that would push me to upgrade to the latest and greatest model. Unfortunately, these little cameras have now gotten to Microsoft Office status, with all of their incremental improvements being about as exciting as new Word templates. Even the big reviewers are getting the message— this week, David Pogue reviews a low end Fuji camera that he claims is targeted at "Generation Z" because it has a number of features that prep photos for eBay, Blogger, etc. before they leave the camera. Overall, it seemed like a very ho-hum proposition, though he does mention one interesting feature: the Fujifilm Z10fd comes with an IR port to blast pictures Palm Pilot-style to other cameras similarly equipped. Though this seems like a fairly borked implementation, I'm sure that this is where cameras need to go in the consumer segment to become exciting again.

As he often does, Dave Winer hit the nail on the head a few weeks ago with his concept of the "social camera" where the whole point of the camera's design was to enable the sharing of pictures. I think he hits the nail right on the head in that I can not count the number of times that I've asked for/been asked for pictures tha somehow never make it out of the camera/PC of their owner. And unfortunately, just about all photo-sharing sites (Flickr being the one that comes closer to being the exception) are far too much of a band-aid solution, mostly because of how labor intensive the upload/tagging/sharing sequence is.

The right solution starts from a network connected camera; however, this is a necessary but hardly sufficient precondition. Having had a Nikon S7c, and having been a beta user of the wonderfully clever Eye-Fi wireless SD card, I can attest to the fact that poorly implemented network connectivity can hinder the experience in two key ways. First, if it is too hard to get associated with an wireless access point, and too manual to pick the photos one at a time to send to the cloud, it just won't happen (this is where the Nikon really fell down). And second, if the camera doesn't connect to and send the pictures to a very open cloud-based service with an extremely simple and well-documented API for extracting them for inclusion into other services, there won't be the opportunity for other people to write innovative applications around the photostream.

In fact, if I was a camera manufacturer, I would do two things: first, I'd look at the way that the iPhone works to get on and off of wi-fi networks to see if I could do the same. And if the processing power was just too much for the electronics in a camera, I'd get a Bluetooth module on board and piggyback off the fact that everyone carries a phone, and that increasingly, most of these phones will have Bluetooth and all-you-can-eat data plans. The second thing I would do is build a river-of-photos service backed by Amazon S3 for all of the photos taken by the camera to go to. The challenge would be handling privacy, but with a little thought and some smart defaults, I'm sure something could be worked out (or perhaps early implementations of the system could be just for "Generation Z" who doesn't care about sharing everything).

The closest I've seen anyone come to this is the Shozu series-60 application that would take all of the photos I would take on my Nokia e61i and upload them to Flickr in the background. When I retired my Nokia, Shozu was the thing I missed the most, if only because it hinted at the possibilities of managing my photos as one big feed, with all of the same tools and tricks that I have come to increasingly use with all of the other feeds in my life.

Are you listening camera vendors? I'm ready to buy if you're ready to make!

Comments Permalink

Print 2.0 observations

The show was a fascinating— if exhausting— mix of partners, customers, and analysts. The sheer number of products and services introduced by all parts of IPG was overwhelming, so rather than focus on them, I'll just stick to a couple of the observations that came from my meetings with partners and customers and the various conversations in the hall.

Content is getting smaller and smaller: whether I was meeting with content owners or retailers or even just analysts covering the enterprise space, there was a common theme around delivering "just the bits that matter" to users with the implication being that in the past, people had to settle for a more shotgun-like approach to getting the catalog, the newspaper piece, the map, etc. Some of this apparent bias may be because most of these conversations took place in the context of the Tabblo Print Toolkit (which lets a content owner extract and format only the relevant chunks of content), but overall, thinking blog post instead of article, 20 second clip instead of infomercial, and widget instead of website seems to be something that HP's partners and customers are all hip to these days.

Platform is now undefined again: a platform used to be what you built a house on. Then the computer revolution happened and platform became the name for something desirable if you wanted to make mucho money as a vendor. It generally implied lots of up-front R&D amortized over hundreds of products and millions of users— all managed super carefully to yield a closed but self-reinforcing ecosystem. I'm not sure what platform means anymore, but I'm positive that it doesn't mean that. In fact, I was surprised to see that even for those HP partners who pioneered and mastered the platform strategy as defined above, the concept has taken on a whole new meaning. Some people use is to mean community, some use it to mean value chain, and others still use it to mean something between infrastructure and protocols.

I suspect that when real definition emerges it will contain elements all of these qualities but will differ from its preceeding definition in an important way— users will be playing a much more active role. Whether it is a printing platform (a couple of which were introduced this week), a content platform, or an enterprise service management platform, it was very refreshing to see VJ and team embrace the ambiguity of the redefinition and welcome all forms of participation going forward with partners, customers, and even the general content-consuming public.

Now it's time to get back to executing on this whole thing...

Comments Permalink