Hacker News Reader

Comment by michaelhoffman | Hacker News Reader

As someone who works in genomics, all these issues apply to genomic coordinate systems as well—same concepts, different words.

View on HN

And as someone who works in CAD/CAM/robotics software, these issues apply to robotic coordinate systems, too: World coordinates, joint coordinates, polar coordinates (not the same as high-latitude GIS coordinates, think angle + distance instead of X and Y), tool coordinates, tool offsets, work offsets - all of those are just vectors of 3-6 numbers. Instead of esoteric international standards, and huge modern software packages, we have a lot of old machines, a lot of old code, and a lot more tendency towards 80s-style strategies: Post a sign stating "Do not dumb here!" and leave it all up to the operator to get it right.

There's nothing like telling your 6-figure CNC or multi-axis robot to make a piece of carbide occupy some space where you currently have precision-ground hardened steel because someone left a G92 somewhere you didn't expect it to be. Though I suppose that errors can be at least that expensive in GIS if you're quoting civil engineering or real estate...

As someone in computational physics, let me assure you it's the same there. The one thing is doing math right and having the correct idea about coordinates. The other is being able to tell the computer about these ideas! Many people cannot code, so are not aware of array or vector facilities of their programming language/ecosystem they use (or rather have to use, because it's always have been like that).

This Phd comic is pretty acurate about the situation: http://phdcomics.com/comics.php?f=1689

Now imagine being a newcomer with neither programming experience nor domain specific knowledge and asked to add some feature to such a legacy codebase. This is the standard case in science.

I feel seen :-)

In my first job I updated someones (doing a PHD in fluid mixing) Fortran code.

The original software to run an experimental rig just put up a ? prompt and you typed numbers in to drive the experiment.

You had to memorise which number was used for what and at which point you had reached.

I updated it to prompt the user and at every step display what options where available and also allowed you to go back a step.

I even went as far as to use Mixed Case Hollerith segments!

My thesis supervisor was surprised when I handed over my code in git repo, CDROM and USB drive.

Apparently they're used to just receiving print-outs shudders

Reply on news.ycombinator.com