FOSS GIS is incredible these days. You can do everything in open source - it just sometimes requires a bit of work. Between qGIS (+SAGA/Grass/etc.), postGIS, python tooling like GeoPandas, leaflet/cesium, etc. there's plenty of tools for different things outside the ESRI ecosystem, you just have to learn to use them. This is a long way from the dark ages 10+ years ago when much of this capability was closed source or via some ridiculously complex text interface of GRASS.
ESRI's market dominance on GIS analysis is still a big thing though but only really a hangover from the days when ArcGIS had much more capability and lock in coroporate training was a thing. Now qGIS has essentially most of the same capability, but ESRI's corporate training structures etc. are still there to some extent. It is losing its stranglehold though, especially amongst GIS 'dabblers' who don't want to sign up and lock in to a massive corporate account for whom the ESRI older-style value proposition isn't really there. qGIS is even an accepted option for local government these days in many places.
Geospatial is both complex and diverse though. There's a lot of industries there and while there's big easy simple 2d GIS markets in mining, government zoning and military, there's also a lot of other stuff going on and its all moving fast. 3d, temporal data, drone photography, lidar and satellite etc. There's lots of innovation happening there right now and new ESRI monoliths being built.
For example, few consider going with Oracle anymore now that you have PostgreSQL but that took a long time. So, looking perhaps as far as 10 years, which FOSS GIS software might even have a shot to start displacing ESRI in the future?
- Web - ESRI never had a major horse in this race. Google Maps is still there but more common now is Leaflet and Cesium for 3D is basically the only option. Both Leaflet and Cesium are FOSS.
- Server - As you say postGIS is now the industry standard
- Programming - There's some people using ArcPy if you are in the ESRI ecosystem but the plethora of FOSS libraries out there are probably bigger in scope and fit together better than ESRI. PyQGIS is another option but in my experience not as heavily used as other open source libraries (all of which are based on GDAL, OGC standards and the Java Topology Suite)
- Desktop analysis - This is ESRI's core stranglehold and I think qGIS is now a competitive option for most users (certainly it has most of the core functionality) however there's a lot of inertia to move from big companies especially those that want ongoing external support/training.
Reminds me of the opioid industry!
I could rant about sas for ages. Its like it was purpose built to encourage anti patterns and Its own docs show code style is a foreign concept to the company.
This, a thousand times over. During my first geography degree I had a lot of GIS modules. I loved it, the capabilities it offered and so on. But it soon became clear that if you wanted to become genuinely good at it, you needed to learn programming, or at the very least have a very basic understanding of it. In this respect, my course was woefully inadequate; professors would perhaps touch upon it rather tangentially, and my requests to include programming in the course modules were brushed off. "You'll do more of that during your Masters!" they said. The reality is that even the MSc had very little of it. If I could say something to all those fresh-faced GIS undergrads, is this: learn programming; it will give you a much better understanding of how it works, it will greatly increase your capabilities in terms of what you can do (as Waterluvian so eloquently said).
For those of you wondering, I did my undergrad in the UK, graduated in 2009.
Also, greetings from a fellow geographer!
Challenges presented in spatial data are really really fun, and interesting, you can extract so much more information if you also include their spatial relationship.
Sadly the industry itself is severely underpaid. If not for my excellent colleges, challenging but fun problems i would swap to typical soul crushing webdev job for thrice my pay.
TL;DR: Lean programming, and also learn GIS - they work perfectly together.
It was never meant for that. That is/was Google Earth Pro.
Not to be confused with Google Earth Engine, which is the most shocking piece of technology I’ve come across in my career and is absolutely free if you can get access: https://earthengine.google.com/
Earth Pro = prosumer Google Earth.
Earth Enterprise = private, offline globe.
Earth Engine = research-oriented massively parallelized ML-based imagery analysis platform.
Echoing your point though, you have to be willing to code. However, I've been working with geospatial data for going on four years now in various academic settings, and it still blows my mind that I can process a very large amount of publicly available data, very quickly, on old equipment and make meaningful contributions to research & policy.
Source: Worked through ~1.3K hdf5 files in the last couple of hours
https://roscon.ros.org/2018/presentations/ROSCon2018_Unleash...
FOSS GIS software is pretty good but suffers from the usual lack of budget. Nevertheless, I’ve been able to push robotics + GIS packages for QGIS with ease. Reaching out to my ESRI contacts about “where’s your indoor and outdoor robotics suite offerings” received a response of only crickets.
As for Google Maps. The biggest tell is that there’s almost no geospatial capability in the software. They don’t want you to analyze and understand. They just want you to consume “where is X”. It’s a standard Google vector for ads approach.
I’m going to rant for a long time if I don’t stop here so I’ll leave it at: stop teaching so damn much ESRI and start teaching way more programming. The biggest regret of my education was that I was mostly taught how to use Geospatial tools rather than how to make my own.