Wednesday's Child

by Kurt Cagle

When all software is free, what is the value of software? Is it worthless, or priceless? In hindsight, it is pretty obvious that people like Kevin Kelly and others of that ilk were very wrong about the trends, where ultimately everything would become free and a grand new society of plentitude would emerge. Gas, while dropping down in the last few weeks, is still trending upward, food is becoming dearer by the day, the cost of health care services is rising dramatically, and indeed the costs of most raw commodities - trees, metals, oil, and so forth - has been aggressively pushing upwards, and will likely continue to do so, though not without a few corrections here and there.

No, ironically, it is only those things that are fundamentally intangible that are diminishing in cost - software, movies, music, the value of the US dollar, that sort of thing. Finished goods are reaching a critical breaking point - they aren't selling anywhere near in proportion to the amount being produced, but at the same time the cost to produce them continues to rise in proportion to both raw material costs. Labor costs still factor - in some areas there is something of a labor crunch going on, yet the crunch is nowhere near as grave as it was in the late 1990s, and at least anecdotal evidence seems to indicate that this is easing except in limited spot markets.


2006-09-21 03:47:31
The good news is that if you have XML skills ++, the commoditization is favorable because in some of the market domains, that and ASP.Net make one employable. On the other hand, both of these skills are orders of magnitude easier to acquire than programming skills from the olden daze of the programming priesthood.

When a market enters a lagrange point, selections that affect the basin of any object in motion are related but not necessarily strongly coupled. For example, hiring decisions are made based on near skills such as advanced degrees in mathematics or on temporal trends such as targeted selection interview techniques (I'll blog that one later today).

The demand for software hasn't diminished. The cost of distributing it has. The one hole in your analysis is you are looking at commercial software that is distributed. Just as in music where movie scores are more lucrative than pop songs and far less risky, works for hire, works for limited distribution (eg, embedded systems, military applications, high-end manufacturing and plant design, etc.) remain profitable. All see falling costs but if there is a sustainable demand, a barrier of limited subject expertise (think middle tier objects), and a lifecycle replacement time where replacement is frequent enough such that saturation does not reduce profit margins below supplier costs for sustaining the production. commoditization still increases the market size while reducing unit production costs.

The winners in these markets reduce the decision cycle time, so fast effective analysis is vital for the winners. What you are seeing is another round of re-engineering of the businesses, reduction in overhead costs, and real winnowing of the dumber managers or the inexperienced ones who thrive on pop management books (eg, The Tipping Point). Experience and hard pragmatics count. Chasing trends and pushing personal political agendas (eg, smashing the glass ceiling) have much less value. Beware analysis paralysis. It's better to lose a few product opportunities than to lose a whole company.

Kurt Cagle
2006-09-21 21:35:45

I think your analysis is spot on ;-) though I do have a few points I'd argue.

XML is not, in general, significantly easier than programming in the bad old days. Anyone who has had to wrestle a complex schema (or more complex ontology), anyone who's had to deal with monads within transformations (besides possibly Dmitri Novatchev) or dealt with asynchronous distributed systems is probably just as proficient a programmer as someone who wrote core assembly routines - their proficiencies simply lie in very different areas. I'll be the first to admit that as a C++ developer I'm at best marginal, not because I don't understand the language but because I have not spent the years of exposure that make that understanding second nature. However, I can reasonably claim to being one of the better XML developers out there (not the best by a long shot, but certainly proficient).

I have to admit the idea of thinking about the hiring process as a Lagrangian analysis is intriguing. If you look upon the various attributes of both a job candidate and an employer as fields (either attractive or repulsive, with appropriate scalar and/or vector characteristics), then the orbital mechanics (in n-dimensions) of the relation could in theory be plotted to determine the best fit , i.e., which candidate is most likely to remain in a close orbit to the employer, which are likely to be unstable relationships and which will never interact.

I am not so much looking at commercial software as that software which will keep a developer renumerated. If I build a super open source mail protocol and application that becomes wildly successful and makes it in the best interest of a company to keep me "employed" as a consultant in order to increase their own reputation or keep the expertise close to home, then this is roughly equivalent to being a jobber working on commercial ERP software.

My guess is that the "superstars" in programming will tend over the long term to end up more in the above position - think Torvalds at OSDL for instance, or the many big XML names currently working for Sun or IBM as "evangelists" or "technical researchers". Whether they have immediate value to the company as programmers is debatable, but because they are considered experts in their respective fields they also enhance the reputation of the company and attract other programmers that view these people as icons.

-- Kurt

2006-09-22 03:15:13
I agree that some applications of XML are quite difficult, but overall, XML itself and the many framework supports (eg, having XMLhttp) make it easier to learn than the days when the first question was do we count from left to right or right to left. There is a lot more that is freely available.

Interestingly, the company that wrote the literature I am reviewing talks in terms of 'dimensions' and 'stars' which are concepts for eliciting behavioral predictions based on past behaviors. Many model-oriented systems of analysis that assign scores over multiple questions and interviewers that are then combined for consensus probably are lagrangian. So are systems that use multiple XML languages to create final renderings particularly real time dynamic renderings. It's a fun intellectual model when one considers the various techniques such as low-transport vs Hohman transfer being applied to time/energy optimizations in evolutionary stable systems. One could think of targeted selection as a kind of ESS.

Right. Big score software (publish it, make big money, take the tours, talk at the conferences, get a house gig as the main attraction) is different from acquire chops, learn the set list and get a house gig as a sideman (eg, team programmer in an IT shop) are different markets. My only point is that the first kind of gig is definitely on the wane (saturation) and the second kind is always there and decently paid. It does have a missile gypsy quality to it though. It helps to be young and portable. As far as the company fortunes, it is still a good business but a lot harder and more competitive and that puts pressure on the economies of the markets.

Tucker Petersen
2006-11-26 03:20:25
Singer George Michael lends the piano on which John Lennon wrote Imagine to an anti-war exhibition...
2007-04-14 23:11:14
Interesting comments.. :D