Thursday, August 11, 2016

You Can't Go Home Again


Just last month, Google won its trial brought by Oracle over the Java APIs it's owned since it's acquisition of Sun Microsystems in 2010. This week Hewlett-Packard Enterprises announced it's acquiring SGI (formerly Silicon Graphics).

The closely occurring events caused me to reflect on the computing technology landscape I've trod through over the last 30 years.

Like most physics majors, I started out squarely in hardware-land (Motorola 68000-land, actually).

I didn't start any serious developing  until I was teaching comp/sci in the mid 80s.  Up until that time I had mostly worked with large systems in academic environments (Xerox and IBM).  But as a teacher, I had lots of presentation-oriented requirements for what I wanted to do.  That led me to Commodore's Amiga systems which were 68020-based, had a custom graphics chip-set, full multitasking, and stereo sound. In 1985, there wasn't a lot out there  to compete with that at reasonable cost.

So I became a developer with C and Forth on AmigaDOS.  Had a lot of fun for a few years before the problems with the company, not the platform, became impossible to overcome.  We developers used to call the Amiga "the Computer Commodore couldn't kill". Technically, they actually didn't kill it, though they lagged development and were eventually overtaken by PCs in performance.

By the early 90s it became obvious that Commodore was dying a slow and painful death. Changed horses in the middle of the stream. After surveying, the small platform landscape, I decided to jump ship for Steve Jobs' NeXT.  Based on Mach UNIX, OpenStep was a powerful operating system, and the platform was specifically aimed at being an academic workstation.

But development was very frustrating as promised compression hardware failed to show up, and the company made frequent hardware changes without notifying developers.  And, of course, eventually, the software platform was absorbed into Apple as was Jobs.



By this time, I was doing a lot of scientific development and landed in a space dominated by Sun, SGI, and Cray. Lots of intermixing going on there.


  • SGI ate Cray and SGI sold off it's  superservers to Sun. 
  • SGI spun off the Cray Research Business division to Tera.
  • SGI went bankrupt.  
  • Sun was eaten by Oracle.
  • SGI was eaten by Rackable. Rackable changed its name to SGI
  • SGI returned as a much smaller player
  • Cray returned as a much smaller player
  • IBM started to dominate the HPC and supercomputing space.
  • SGI got eaten by Hewlett-Packard Enterprises.


Had fun, but none of those environments are going to be significant in scientific computing in the future (for now).

So I can't go home again.  The whole neighborhood has burned down.

Most of my computing today is centered around machine learning applications.

The dominant platforms are being provided by IBM and HPE.

This old dog  may once again have to learn some new tricks.  Or maybe just retire...



You Can't Go Home Again


Just last month, Google won its trial brought by Oracle over the Java APIs it's owned since it's acquisition of Sun Microsystems in 2010. This week Hewlett-Packard Enterprises announced it's acquiring SGI (formerly Silicon Graphics).

The closely occurring events caused me to reflect on the computing technology landscape I've trod through over the last 30 years.

Like most physics majors, I started out squarely in hardware-land (Motorola 68000-land, actually).

I didn't start any serious developing  until I was teaching comp/sci in the mid 80s.  Up until that time I had mostly worked with large systems in academic environments (Xerox and IBM).  But as a teacher, I had lots of presentation-oriented requirements for what I wanted to do.  That led me to Commodore's Amiga systems which were 68020-based, had a custom graphics chip-set, full multitasking, and stereo sound. In 1985, there wasn't a lot out there  to compete with that at reasonable cost.

So I became a developer with C and Forth on AmigaDOS.  Had a lot of fun for a few years before the problems with the company, not the platform, became impossible to overcome.  We developers used to call the Amiga "the Computer Commodore couldn't kill". Technically, they actually didn't kill it, though they lagged development and were eventually overtaken by PCs in performance.

By the early 90s it became obvious that Commodore was dying a slow and painful death. Changed horses in the middle of the stream. After surveying, the small platform landscape, I decided to jump ship for Steve Jobs' NeXT.  Based on Mach UNIX, OpenStep was a powerful operating system, and the platform was specifically aimed at being an academic workstation.

But development was very frustrating as promised compression hardware failed to show up, and the company made frequent hardware changes without notifying developers.  And, of course, eventually, the software platform was absorbed into Apple as was Jobs.



By this time, I was doing a lot of scientific development and landed in a space dominated by Sun, SGI, and Cray. Lots of intermixing going on there.


  • SGI ate Cray and SGI sold off it's  superservers to Sun. 
  • SGI spun off the Cray Research Business division to Tera.
  • SGI went bankrupt.  
  • Sun was eaten by Oracle.
  • SGI returned as a much smaller player
  • Cray returned as a much smaller player
  • IBM started to dominate the HPC and supercomputing space.
  • SGI got eaten by Hewlett-Packard Enterprises.


Had fun, but none of those environments are going to be significant in scientific computing in the future (for now).

So I can't go home again.  The whole neighborhood has burned down.

Most of my computing today is centered around machine learning applications.

The dominant platforms are being provided by IBM and HPE.

This old dog  may once again have to learn some new tricks.  Or maybe just retire...



Thursday, July 28, 2016

Today's Practical Math Problem

Presidential candidate Trump says Mexico "will pay for the wall".  He plans to amend the Patriot Act on his first day in office to require that "no alien may wire money outside of the United States unless the alien first provides a document establishing his lawful presence in the United States". If not, he'll withhold remittances from Mexicans living in the US, a sum estimated at $24B annually. He will forego this if Mexico agrees to a payment of $5-10B to fund the construction of the wall.

Note: Conservative independent estimates price the wall to at least $49B. I find that reasonable.  With current concrete costs coming it at an average of $93/cu.yd, making even a 1-foot thick wall would cost at least $3B not counting labor and property acquisitions. And of course the wall would have to be more than one foot thick.

Our GDP is about $17T. Mexico's is $1.14T. Mexico's federal budget is about 28% of GDP or about $320B. (We both run 3% deficits). $49B would be about 16% of the entire federal budget.  Mexico would either have to borrow and raise the deficit to 4% or raise taxes which the government has already promised not to do.  Average annual income in Mexico is about $17K.  (The enmity this would cause would be a good anti-US platform to run on in the next Mexican election).

Of course, President Trump is letting them off lightly with only demanding up to $10B or 3 percent of the annual budget. 

The candidate has not clarified whether Mexico would be allowed to break up the payment into installments over time or whether the US government would pick up $40B difference. The more time involved the more illegal immigration will continue unabated.

$40B would be more than twice the requested budget for NASA in 2017.

1. Would it be more correct to have Trump supporters chant Mexico will partially pay for the wall?

2. Won't Mexico have a significant amount of time to consider their response since the President can't amend the Patriot Act by fiat.  A bill will have to be submitted to congress and passed by both houses which could take time.  I imagine if Republican majorities in the house and Senate are maintained or expanded this could be accomplished. Of course nothing takes place in a vacuum in politics.  I presume companies involved in profit from remittances such as Western Union would be lobbying congress heavily against the amendment.

3. If Mexico pulls out it's wallet and pulls out a $10B bill, meeting the Trump request and assuring continued remittances, will our taxes have to be raised or government services cut to cover the remaining costs? 

Note to small government conservatives: This would not result in an overall lowering of government expenditures. Just a shift of what the funds are spent on. 

4. Building the wall would result in a boom in the construction sector in the states bordering Mexico,  lowering the overall unemployment rate.  (President Trump could legitimately argue that he has brought blue collar jobs back to at least a part of America). This boom would be temporary since at some point the wall would be complete.  The Trump campaign has released no estimates on what the ongoing costs to maintain a nearly 2000 mile wall in perpetuity would be.

Note: An interesting and ironic turn would be to charge each of the border states with maintaining their section of the wall.  This would keep with conservative philosophy that states should handle their own business as much as possible.  One assumes that during an upswing in the national economy, these states might come to resent the additional financial burden.

5. We also share oceans with Mexico. If the wall precludes land-based migration, is it reasonable to expect an uptick in coastal migration? If so, would that require additional federal expenditure on the US Coast Guard which would also be expansionary to the budget?  Should the Trump campaign modify it's request to Mexico to include these likely costs? Or modify it's tax plan to address this?

Thinking through these economic questions is a useful exercise in understanding trade and economic systems regardless of your political philosophy and party affiliation.

--C--

"I taught econ for a couple of years"



Thursday, February 4, 2016

President Bernie and the Supremes


At the February 4th Democratic Debate Bernie Sanders stated that  he would not appoint any justice to the Supreme Court who hasn't spoken out loudly that he or she would overturn the Citizens United decision.

That would be a big change from the norm wherein most nominees to the court are extra careful to NOT indicate how they would vote on an issue prior to being appointed.  This sort of thing was frowned upon after the Bork debacle during the Reagan presidency.

A President Sanders would most likely have to get his candidate approved by a Republican-led senate which I would think would make such a nomination dead-on-arrival.

If Sanders COULD get someone who both passed his litmus test AND successfully navigated the congressional consent process, Citizens United wouldn't be reversed on the first Monday the following October.  Someone would still have to bring an appropriate case in the lower courts and that case would have to progress all the way up the federal court system, which means an awful lot of appeals (translate: It would take years).  Then the Supreme Court would have to agree to hear the case, and the case would have to have broad enough enough ramifications to effect the Citizens United decision and not have a narrower interpretation.

As a sitting senator, Bernie Sanders is obviously aware of how our system works so, considering all it would take to get that decision reversed, I can't help but find Bernie's statement disingenuous.

Not hatin' Just sayin'

Subscribe Now: Feed Icon

About Me

My photo

If you're looking for the Court Cannick who's architecting ComputeSpace for education or the Court Cannick who built NetworkMathematics for systems management or the Court Cannick who engineered the storage systems for SETI@Home or the Court Cannick who produced Physical Science Journal for Storer Cable or the Court Cannick who lectured on Space Colonization at CMSI, then you probably have found the right Court Cannick.  Otherwise, keep looking...