Those of you that know me and live in Milwaukee know that I'm in the process of starting a software firm. For this venture, IronWake, LLC, I have a great business partner, Pete Reynolds, who has been a colleague and friend over the years. I trust him. He is a great guy and has strengths in business that complement (there's that word again, Gerry) my skills. He is a good marketing guy, a good product manager and is focused on many aspects of this firm. I am squarely focused on building product and understanding the sales channel. I've been in a few startups and have seen that a lack of focus in sales, or in other key areas, means certain death. I'm not afraid of failing, but rather will focus on spending the time and energy necessary on the things that make a firm successful. It'll be interesting over the next few months as we ramp up and develop this firm. I'm looking forward to the day when we have a professional staff, i.e. a CEO/COO, a sales team of more than two people (no, not including Pete and I), a marketing team again of more than two people, customer support teams, development teams, etc.
So, Pete and I are at a fundamental crossroads. Do we raise capital in the form of angel monies, or do we bootstrap this operation? There are a number of key factors that we are considering in this process. Pete and I are unafraid to give away equity or do a deal involving some form of equity, debt or mezzanine financing, which may require us at some point to relinquish as much as 60% of the firm's equity over a few rounds of financing. No big deal if we are making dough throughout this whole process. Pete and I see eye to eye on all of these issues. We want to make cabbage and there are plenty of impediments between us and that goal without the two of us and our eventual partners tripping over each other and mucking up the process. It's all about FOCUS and SIMPLICITY.
I'll not get into the details of our software product at this time. I also want to be clear about my mention of IronWake on this blog. IronWake will likely be a bigger and bigger part of my life as time progresses. I'll mention it from time to time here, but I want to be clear. I'm interested in sharing my thoughts about this firm and the development from a professional developer/designer/architect perspective. If I market IronWake here, it is purely out of habit. If I talk about IronWake and IronWake products, it is because I feel that I've found something interesting to share with anyone who reads my junk here. OK, stare at this spot:
.
.
.
You are getting sleepy. BUY MY STUFF.
----
If you want to contact me, feel free to do so here. If you'd like to preorder an IronWake product, sorry, but you'll have to look elsewhere. I won't link it here.
Dr. Don Syme and his good group at Microsoft Research have released F# 1.1.0.4 (with pre-release notes here). Way to go guys. While I have a copy of this software, I haven't had a chance to work with this release yet. I currently am using 1.0.8.6 and expect to upgrade a few VPCs to 1.1.0.4. I didn't get a chance to do this over the weekend as I had hoped, but I have looked at the docs and samples. For those that don't know of Dr. Syme, he was, among many other things, an early champion of the Generics that we now see in C# 2.0. His list of accomplishments is long and as I cannot do this list justice, take a moment to look at what he has on his home page.
I'm a math guy that programs (or maybe vice-versa), but it took Marty to point me in the direction of F#. I had heard of F# some time ago, but had mistakenly thought that it was the replacement for Fortran in Microsoft's suite of developer products. Silly me. Marty mentioned it about 8 weeks ago or so and I have been hooked ever since.
Given that GAI.NET is way behind schedule for release, I'll not make any promises here. I hope to release some community projects on F# in the coming months, but I'll leave it at that for now.
I do considerable math work in .NET and other mathematics software packages, so I'm optimistic about F#. The language has the feel and syntax of ML and Ocaml. Dr. Syme and his team have added significant functional programming features to the language in this release as well. While I don't like to prognosticate on these matters, I do see that the ability to use a rich language that is mathematics friendly will be critical for the .NET practitioner when doing numerical analysis, scientific computing or other similar disciplines. The fact that this is "pure .NET", i.e. it is interoperable with all of the other .NET languages, and is compiled means that over time, I can invest in F# without concern for loss in value of my time. As F# is still a research language, this will be over a long period of time, but I can at least start the planning and proof-of-concept on my mathematics infrastructure. Given that the most expensive cost in a typical infrastructure is the "shuttling" of data amongst the computation engines, I expect that I can take a more comprehensive approach to dealing with data. I mentioned to Scott C. Reynolds (I'm working on that name thing, I need more material) recently in a comment here that I am planning for C# 3.0. This is precisely how I think about these problems, given that C# will make use of support for data in the language. Using discriminated unions in F# also improves much of my ability to deal with these issues as well. My toolbox just got a lot bigger.
One other feature that is crucial to what I do is "big integers and arbitrary sized rational arithmetic". I do a considerable amount of Number Theory and had to move to other platforms to get this functionality. SCORE A BIG PLUS FOR THE F# GUYS ON THIS FEATURE. As time progresses and this language matures, I'd suspect that many of the features that are available in other environments such as Maple, S+, MatLab and Mathematica will be core/standard features of this language. The features that are not placed in the language will likely be simple to implement. This means that Graph Theory, Algorithmic Number Theory, Cryptanalysis, Cryptography, Genetic Programming (GP on steroids in F# is an understatement), Emergent Computation and other assorted work have a new place to live at JJBR and The House of Scalper. I get all of the benefits of the .NET runtime environment with distributed computing, etc. as well.
So, to Dr. Don Syme and his hard-working team, I salute the work that you do to further this language. It is sincerely appreciated.
The last few weeks have not been kind to me. I have had a 200GB HD failure (part of a 1TB raid5). The reason that this failure is significant is that I had another identical drive failure on another of my 1TB raid5 about three months ago. Since I used my last free drive to swap in on that raid and hadn't replaced my spare 200GB drive to cover another failure, I found myself vulnerable on the second raid. Who would have thought that two different Western Digital drives would go bad within a three-month period? Gee, I would. I've had nothing but bad luck with their stuff and hesitated when I bought this round of storage 18 months ago. While I didn't lose any data or any other collateral, I lost time in restoring hot backups. Since the second raid is vulnerable and I refuse to get anymore WD equipment, it can slow-rot. So I purchased a LaCie 500GB external drive a few weeks ago for temporary storage of the VPCs on the vulnerable raid. I'll either have to cut down on storage, or purchase more. I'll decide shortly, but I'll likely not make more purchases after this because the old budget is getting strained.
Now, I have to say, that I'm Mr. Cautious about my data. I know what I can throw away, I know what I can regenerate, and I know what is critical and needs hot backup on the network. I earned my Boy Scout Merit Badge in Systems Administration (not really, but I'm courteous and kind to my data, so that should count).
So I make copies of my VPCs to the LaCie 500GB drive (this is connected by firewire-800). I leave this drive for about a week while I use the live VPCs (with no significant changes). After sitting one week, the LaCie is powered up and voila, it is unavailable, making a clicking sound and generally making me feel that God hates me and my data (or my data and I as the case may be). Failed drive, lost data, etc. etc.
Now it is off to the computer store to get an even exchange. Now, I'm really mad at this point because this drive is 23 days old from the date of purchase and CompUSA only has even exchanges through the first 21 days. I can't find the original box (somewhere in bedroom #3) and I'm worried that these guys are going to stiff me and leave me to do an even exchange with LaCie by mail. GREAT. I get to the store, wait in line behind a few customers and finally get to tell my story. The guy looks at my receipt, realizes that my drive is out of the 21 day period (and I don't have a box), calls the store manager while he fiddles with the cables. Without a discussion or any argument, he says, "Let me make sure we have one in stock. We have to keep the box of course, but we'll do the exchange." WAY TO GO COMPUSA. Unlike Josh's recent experience with Tiger, I got great service.
The story is not done just yet. I get the new LaCie drive to the house, connect it to the computer (again on firewire-800) and proceed to do a full format (as instructed in the manual). No problem. Roughly three hours later, the format gets to 99% and then outright fails. Ye-haaaaaaaaaa. No problem, I'll just do a quick format. Now I get that silly failure message again, only this time immediately, showing a message something like "Unable to format the drive". My trust in hard drive manufacturers is slowly deteriorating. HEY HARD DRIVE PEOPLE: WORK ON QUALITY INSTEAD OF SIZE. I'LL STILL BUY YOUR DRIVES. I PROMISE. My 12TB of live storage and 8TB of hot backup are testament to that. Get it right you morons. OK, I feel better yelling a little here.
So now I think to myself, hmmmmmmmmm, maybe it's the interface and not the drive. I connect the drive via USB 2.0 and try the format. OK then, we have liftoff. The format completes and Windows is happy with the drive.
I then decide to trust the drive, because I can't have four drive failures in a matter of three months now, can I? I decide to fill this drive up with duplicate copies to test this thing. So far, so good, but I'll let you know in the next few weeks if LaCie gets placed on my bad list. I'm tossing 3TB of data (that I haven't had time to analyze, but that I can regenerate) so that I have extra backups of everything on the LaCie and additionally have hot backups for the failed 1TB raid5. All of the drives in that raid5 get to play the role of spare drive on the aforementioned initially failed 1TB raid5. Again, all of these drives are WESTERN DIGITAL. I'm also running hot backups to another raid so that in case I'm surprised by another WD failure (no I won't be surprised), I'll have plenty of other faulty WD drives for that raid. In the meantime, I've got a full size axe, even though I'm not an outdoors kind of guy, that will be useful in preparing the failed WD drives for the junkyard. We wouldn't want to see those drives make it to the dump in one piece now, would we? There's valuable data on those there drives and there is nothing better than a simple workout chopping wood platters. I can justify my recent health club membership, but there is nothing better than a good old workout in the backyard to take out those frustrations and whatnot.
Now, we're not done with useless time-sucking activities. The Armed Geometer was outstanding with advice recently on Linux and Unix installations. Since I have a background in Unix in the way old days, I needed to find out what would be useful without wasting a lot of time trying 10+ different versions of Linux. HEY OPEN SOURCE GUYS - BASH AWAY AT MICROSOFT, BUT AT THE END OF THE DAY, I KNOW WHAT I HAVE. Thanks again Andy for the advice here, because in the long run, you've saved me at least as much time as I wasted on my hard-drive goofiness. Solaris 10 installed on a VPC. I'll be repurposing at least one dual Xeon server for Solaris as well. To those of you that know me as a .NET guy, no, I'm not moving to the dark side, but rather taking a healthy step here on some serious Number Theory problems on my agenda. Since many of the good class libraries are available on Unix, I don't want to waste any time porting the software to .NET. Besides, I'm a strong advocate of F# which I'll address in my next post.
Time-waster item three here is finally a problem on my primary workstation for DirectX development. I've been having problems getting the August, 2005 DirectX SDK on this machine. I have .NET 2.0, VS2005 Beta2 and am waiting to resolve the DirectX issues here, as I'd like to get WPF/WCF on this workstation. Again, I have another workstation for redundancy here (with a high performance 3D video card as well), but that machine is not .NET 2.0 as of yet and isn't scheduled to move up to that release. I still need some stability. More time wasted on something that should just work. I'm not pointing a finger on this one as I just need to understand why the DirectX SDK is not succeeding in it's installation. I have to invest my time to get a better understanding of what is going on in this scenario.
So, I'm describing all of this here, because this has had an impact on a number of projects in my personal life. My JJBR work has suffered. My IronWake work has suffered. GAI.NET has suffered as well. I'm nearly code complete on GAI.NET, but I cannot commit to making that a priority, as I have to make certain that the things that "earn" are the things that receive my attention first and foremost.
I apologize to those of you that have emailed me asking of GAI.NET's status. I'll keep you posted.
----
Please contact me here if you'd like to yell about hard drives, give me further advice on Unix or discuss GAI.NET.