or in my VI centric little world:
So I’ve decided to go with Django as my 1st python web framework. The documentation is basically spot on but I hit a typically bizarre setup error about not being able to find make files under the python directory so I loaded up Synaptic Package manager and selected Autoconf and then did a search for python and basically selected close to everything and clicked update and it just plain worked. Outstanding!
A buddy of mine just IM’d me about cheap hosting in the US. His requirements are:
- non shared hosting i.e. dedicated server
- a linux distro which doesn’t have package management issues i.e. gentoo / debian / *bsd
- a non astronomical monthly fee ( he needs at least 2 servers from day 1 to build a clustered architecture)
- ssh access w/ some kind of decent web based admin tool like webmin
- ability to do some kind of disc mirroring / heartbeat with drbd or something similar. This is a high reliability application but he can’t afford to buy his own hardware*
I’ve heard good things about ServerBeach but I don’t know. Anyone have any thoughts? Niall? Where’s your box? Kevin what do you use ? (and yes I could traceroute but I think its a good conversation to start)
*I know that’s inconsistent on the face of it. Don’t bug me about it. Bug him (oh crap; he requested anon)
I think the most common questions any entrepreneur has are always about money — how to get it, where to get it, how much to get, etc. And, sadly, I see everyone always make the same mistake. They focus on what they can raise but NOT on what they actually need. Oy. And I’ve done it myself too.
So here’s my little bit of web 2.0 startup advice for the day:
- Understand your own business.
- There is an inflection point for any startup where a linear increase in growth or capacity requires a non linear increase in capital needed.
I’ve thought about this a lot lately. And I mean a lot and I think the best analogy I can make is from architecture. Reading Paul Graham’s Hackers and Painter’s* book 2 years ago at Foo Camp made me think a lot about architecture. And I think the lessons from building buildings are very applicable to software.
Lets say you live in the Bayou and you want to build a dog house. That can be done with a few scraps of plywood, a 2×4 for 2 and Rover’s is just as happy as a clam. So now you get a 2nd dog and you want to make it bigger. Well more plywood and another 2×4 or some additional lathe and you can easily double the capacity of the dog house. No worries. Now your neighbor asks you to dog sit for them. Ok no problem. You make it bigger still and you can hold 3 dogs. Then all your neighbors bring over their dogs. And guess what? Your dog house sinks into the mud. Why? Well you never made a foundation.
That’s what I mean about a linear increase in growth or capacity leading to a nonleaner increase in capital. You could easily double or triple the size of the dog house and rover et al were fine. But when you needed to quintuple it you really had to rip it down and pour a cement foundation or it would just sink into the mud.
Having had lunch with Kevin Burton from Rojo / TailRank a few weeks ago, I worry that he’s making this type of mistake. Now don’t get me wrong TailRank rocks. But as the quantity of data he’s trying to process expands, he’s going to face this issue. Heck FeedLounge seems to be in the middle of it. Controversy. My guess is that WordPress.com / Matt either have hit this point already or will soon.
So I think this is a basic either VC or engineering principle and I’m going to be cocky enough to name it after myself:
Scott’s Law of Funding: Linear increases in growth / capacity will at some point require nonlinear amounts of capital.
I look back at the founders of Google and that giant pile o’ money they took back in the early days ($25 meg from Kleiner if my memory serves me right) and, damn, were they ever right. Yeah they might a lot richer now and less diluted if they had taken smaller drips as they went but what they didn’t do was ignore this. Whether by luck or planning they had enough money to deal with this up front.
An Internet Example:
When I built out the first data center for Feedster, I bought cheap switches (D-Link gigabit switches to be specific) for our internal LAN backbone**. Why? Well switches are complex and I didn’t know enough to buy the right ones. Now we’re using *censored* but its an enterprise level switch. Our traffic may have gone up by *censored* fold but our switch costs went up by *censored* fold but its a nonlinear price increase.
So keep this in mind. For whatever you do, whatever your startup, there will be a point where your growth rates are going to be such that you need a non linear increase in capital. Yes I know you’re a hot web 2.0 startup and you’re going to be acquired before then. Well that’s the plan and that’s always the plan but it rarely actually happens that way. Sometimes you have to actually operate the thing you’re building and, if so, are you going to be able to.
*If you haven’t ready this yet then you need to. If you haven’t re-read it lately then you should. Heck I just did.
**I did however use Cisco for our external facing side.
So here’s the report on the 1st day of the 30 days. Here’s what it looks like on my system:
And here are some details:
- I moved it from the far right hand edge of the screen to the far left. That feels more right for what its worth.
- I’ve seen a few window redraw issues and my windows programming experience says that that’s tied to applications grabbing the wrong screen contexts / sizes and making assumptions that they had the whole display area. Given that I didn’t see any of that before I moved it, there must be some bad fu associated with left hand side of the screen (and that’s likly why Google put it on the right). Still I read from left to right so having it on the left feels right.
- The internal notepad, “Scratch Pad”, is brilliant. All of my machines since 1995 have had a single file, todo.txt, launched at startup into notepad.exe or another editor. Why? Well its my equivalent of a sticky pad / writing on the screen. This is better since its moderately visual and I can always see it.
- So that’s great BUT their own search engine can’t find text you type into it. I started by entering “where doth these notes go” into the Scratch Pad but now 12 hours after I typed it in, its still not indexed. Consistency thy name is engineer!
- Not knowing where the data is stored for Scratch Pad bothers me a lot but that’s just me.
- The little todo list widget is nice. Very nice. I’m the classically disorganized guy who focuses hugely on what’s in front of him so having this is cool.
- The email widget is really, really nice but I don’t understand the filters UI at all. Here’s what it looks like:
- I know that’s hard to see but the idea is that you put in terms of what you don’t want to see in your list of emails. I like that because I don’t want a list of 10,000 messages in here. But what I don’t understand is how to use this dialog. Subject is a text field as is From and To. So do I put the words in the subject field or the Has the Words field. I suspect that what they wanted was a checkbox or radio button here. Recommendation: See Thunderbird’s filters dialog box. While I have many, many issues with Thunderbird, they really did get filters right.
- It seems hard to do filename searches or perhaps it was still indexing my disc. I wanted to hack the iPodder / Juice code a wee bit to help my snaky goodness (Python) and was looking for xrced and all it found was a mail message containing a reference to xrced but not the xrced script.
- I’m hugely biased towards currency in search results. Hey I founded Feedster right? For me currency is often all that matters. If I’m searching for *.doc what I want are the latest *.doc files NOT everything on my blasted hard disc. I’d strongly recommend to whoever writes this absolute gem of a piece of software that they read Gelertner’s papers on the Timestream / Lifestream approach to organizing data. Gelertner rocks.
But overall I still love this little beast. Its actually gotten me to respond to emails which any long time reader knows I generally don’t do.
My oh my. I’m just plain astonished at the rich, frothy goodness that is Google Sidebar / Google Desktop 2.0 / whatever the heck they call it. This thing is bloody well rocking my world. Yes I know that I’m the last person on the planet to try this new thing from Google which came out more than 6 months ago. And I agree, like Inluminent says so well, that there need to be Mac versions of Google stuff. That said, given that I’m a professional search guy, I’ve seen this type of thing before. Verity did it. Fulcrum did it. DTSearch did it. Others have done it. None of them seem to have done the overall issues of local search as well (note — I haven’t looked at Yahoo’s stuff or MSN’s stuff yet; when I have time). Google really nailed this one well.
So here’s my committment. I’m notorious for NOT using new technology. I still copy files with a DOS shell. I use notepad even tho I’ve gotten 25 other tools, etc. I’m going to give Google Desktop 30 days and not turn it off, not minimize it and see how it shakes out. If I have to give it screen real estate for 30 days then I’ll definitely use it.
The clock starts … now!
technorati tags: search
So I’m seeing Niall and Om and “I wanna be like them”. Seriously though. I’m definitely fascinated by podcasting, I’ve got the iPod 60 gig, and I feel that I’m ready to actually *gasp* create a podcast. What occurs to me that might be interesting to the 13 people who read this blog is this : Tech War Stories. I’ve been in technology for so damn long now and worked with so many characters that perhaps tossing out some of those stories like:
- Every startup needs a green football
- The day paul got a gun or “How to become a consultant in 1 easy step”
- Real engineers can explain their code
- Everyone needs a Geek Kit
If this sounds interesting to you then leave a comment and I’ll start working on it. Most of my best stories are actually pre-Internet so that gives them an aura of “Aged Bits”.
Oh and my 4.5 year old, Alex, has agreed to help. He wants to be the Internet’s youngest podcaster:
Ever hear a podcast that just called to you? So much so that you actually called up the Podcaster and said “Thanks Dude”? Well if not then I recommend you do so. What creative person doesn’t enjoy getting thanked for their work? Well, for me last night (post SES sleep woes have made this now “last night”) it was the Om and Niall show from OnPodSessions. Part of it what was good for me was that the two of them just plain riffed on each other in a flavor akin to a Jazz improv session. Part of what was good for me is that it had an inherently high trust factor — I know both Niall and Om fairly well (Niall more so; Om less so). Just as I inherently trust a Steve Gillmor podcast, I trust Niall and Om.
Nice work guys. Thanks.
Having read much of Defensive Design, the new 37 Signals book, today during SES travel, I found my experiences downloading a new Linksys driver for my old WPC11 v3 card interesting. Here’s the download page. The 1st download link is, unhelpfully, for the Windows CE version of the driver. The second is for the Linux driver. Now I’m all for multiplatform and all but put the most common platform at the top. Sigh.