Hacker News Reader

Comment by nobleach | Hacker News Reader

QGIS and PostGIS were my jam when I worked in that space. We were an ESRI shop so Oracle/SQL Server with SDE (topped with ArcGIS) were the official tools. Some of us were always looking for ways to subvert the culture by building tools based on open source stacks.

One of my favorite experiences from that era: We were meeting with a few ESRI reps for some integration stuff. The lead hot-shot was on his phone playing around during the meeting. He was basically on autopilot. The other two folks were working with GeoJSON response convertors. I said, "I built one of those with TopoJSON". One guy said, "I've never heard of it". I showed them how it was much more efficient and used splines instead of points. The lead dropped his phone and said, "I need you to tell me MORE about that". I showed them the service. They invited me to lunch, I politely declined and said, "today's my last day so I have a ton of things to wrap up". I do miss that realm sometimes.

View on HN

ESRI is a scourge on the GIS industry. One of the creators of postgis can say it much better than I can.

https://blog.cleverelephant.ca/2018/11/esri-dominates.html

Sounds remarkably similar to Oracle, or other "too big to fail" things that sell ridiculously priced software to CTOs.

It's not the software licensing, but the independent consulting entourage that's the problem : internal confidential big four accountancy information told to me without any applicable restrictions on my repeating this put the cost of each Oracle database instance deployed in the F500 at over one hundred million dollars, due to the third party make work.

Ed. "deployed"/"deployment" 0400GMT

The educational world needs to stop supporting them.

But it's the other way around. Esri offers so much in the education sector apart from software licenses in a complete package, which you unfortunately have to search for in the FOSS sector.

The business world needs to stop supporting Microsoft Office. The creative world needs to stop supporting Adobe Photoshop. The industrial design world needs to stop supporting Solidworks/AutoCAD. Everyone needs to stop supporting Microsoft Windows/Apple OSX. etc.

I've been telling people for years they should be using GIMP and Linux but nobody cares. At least Oracle's stranglehold seems to be weakening.

Have GIMP developers ever discussed usability? Because it sure doesn't show!

The alternative to Photoshop is that I gave $40 to Pixelmator and now I have an awesome photo editor that does plenty.

Try Krita - FOSS, excellent UI.

You can add matlab to that list.

Yes, that is the apex of their human capital Ponzi scheme. But, you’d have to wait until the current profs who only know esri retire. Or until they put in the effort to learn how to use foss tools. The only motivation they might have, is a strong belief in foss or pressure from students who know esri is a trap.

That’s not a very good take. Esri is not a Ponzi scheme in any way. There isn’t an alternative for serious GIS. QGIS is a good open source tool but it is nowhere near esri.

As a systems engineer who was hired explicitly for a massive migration from 2 disparate ArcGIS Enterprise environments to an AWS one...

Yeah, there's a massive amount of essential features just missing from OSM and QGIS. Like, even the most basic - layers. They don't even really exist in OSM side of things. Whereas if I'm doing travel-time-bands from medical centers, I absolutely must use multiple boundary layers.

Then again, I'm an engineer, not a data scientist. My colleagues (all data scientists) do all sorts of crazy awesome work. Again, unless you're hacking R/python code and piecemeal it together, you're using ArcGIS. Then again, "ArcGIS Notebook" is a dockerized Jupiter system for doing just that in ArcGIS enterprise.

edit: As for me personally, since I'm working in geoinformatics, I've been teaching myself the hows and whats. Im pretty proficient at both ArcGIS and QGis, as I'm trying to learn them side by side.

If I am doing my own projects, I will use QGis. That's not even a discussion. I'm not about to use proprietary systems to lock up my data.

Can you elaborate why layers are needed in travel time band type analysis?

No, because any further details would highlight where I work and specifically what office.

Lets just say the data intersection of: density and geocoding of users, density and location of office sites, travel time bands, and service level expectations all have a part in the final calculation.

The data is either base layers imported, or calculated layers. There is no one "comprehensive layer" nor does that concept make sense. Different groups handle their own data. We import it in and use their data for our final result.

(In fact, its the OSM approach of everything as 1 layer and tags everywhere is why a DB import and data handling is so obnoxious.)

What features do you think are missing in QGIS which hamper you in your work? I am asking because I am developing a spatial query engine which will work directly with openstreetmap.

I’m not sure that I believe you know what you’re talking about. Not trying to be offensive in that statement, but “can’t back up my very broad, generalized statement with any detail or explanation” comes across as though you’re afraid to reveal your lack of understanding about the details you elided.

You’re saying he doesn’t know what he’s talking about? He is saying you can’t really do GIS without multiple layers. This is basic knowledge for anyone working in GIS.

I am not claiming to be world class expert in GIS and English is not my mother tongue. This comment by tarotuser is confusing me "Yeah, there's a massive amount of essential features just missing from OSM and QGIS. Like, even the most basic - layers. They don't even really exist in OSM side of things. Whereas if I'm doing travel-time-bands from medical centers, I absolutely must use multiple boundary layers." Layers have existed in QGIS for a long time. What are those "massive amount of essential features just missing from OSM and QGIS"?

Explain how qgis doesn’t allow multiple layers. I have a project open right now with 6 layers. That to me seems like “multiple layers.”

And how would that concept even apply to postgis? A postgis query can operate over many tables.

And there are a variety of open source tools that allow multi-dimensional data structures. That’s multiple layers in multiple dimensions.

So I’m hoping, if there’s some aspect of a GIS layer I don’t understand, he can explain it. Qgis totally supports multiple layers in the sense that most people think of it.

He doesn’t have to divulge “secret information” to do this. If he is referring to some aspect of the qgis architecture he can just say so. Qgis is open source, their GitHub repo is clean and well documented. Im merely asking for him to explain why it handling multiple layers is somehow not really handling multiple layers.

Guys I think the pronouns are causing the confusion :-) He who? The blame causality is all mixed up in this thread.

What you’re describing sounds like it fits perfectly into PostGiS. You can have tables and you can normalize your geospatial data. With SQL, and temporary tables, you can build absolutely any analytics.

I worked in a GIS Lab with an ESRI endorsed all Arc stuff textbook author/instructor for a number years. Working on similar projects, for anything remotely complicated, I could implement anything he could do, with some python and PostGIS SQL, usually faster, and instantly reproducible.

ArcGIS is a crutch for people who can’t and aren’t willing to program. This goes for all data manipulation tools that aren’t focused on visualization.

If you have an ETL pipeline, probably best to get GIT involved.

Couldn’t you do travel time bands with postgis? I think so, but if it’s not possible please explain why.

They explicitly target the universities as part of their business model. That’s why I say it’s a human capital Ponzi scheme. They give products out to students and educators way below cost, with the specific intention of making the field as a whole dependent on their products. This is well known and pretty explicit. If professors and students had to pay the same costs as private enterprise for esri products the demand for their products would plummet. This is their business strategy and it’s well known and more or less undisputed.

Whether or not they are a good company that provides quality, unparalleled products is not a point I’m contending. They are not competing on that alone. That’s my point.

What is the alternative for serious GIS? Why should they stop?

That author doesn't say Esri is a scourge; more like it's just generally bad to have only 1 provider in a space -- and that it's up to customers to change that by voting with their feet.

I think Esri is (and has been) in a very similar position to Microsoft's in the late 1990s -- having achieved market dominance, they feel like open source software is the biggest threat to their business. But I think the presence of QGIS is creating competition, which is nothing but good for the industry.

(I worked as a developer for Esri for 15 years)

ESRI have similar network effect to Microsoft. No individual product is that great but taken together it is very competitive. And ArcGIS online is developing that network effect in a similar way to Office 365.

Also, the GIS space tends to have crossover between different departments in companies. You have a mix of IT departments, GIS specialists outside of IT departments managing infrastructure, data management, casual users etc. Each has a different budget and a different competitive environment. ESRI sits as a compromise between these different groups. Each group may be able to adopt a competitor but not gain traction outside their group.

Not having done much in terms of GIS work, I never had to deal with ESRI until last week.

I was on a call to ask some ESRI rep to add some labeled points to my client's existing map tool. It was a somewhat surreal, weird experience where I got the feeling they were making the work seem much more difficult than it was. Their estimate turned out to insanely off the (my) mark, at eye-watering hourly rates.

At first I thought they had enough business and didn't really care about us. But reading this thread, it seems I was wrong. They are the Oracle of the GIS industry.

Adding labels to a map in a readable fashion is not an easy thing. ArcGis solutions, last time I tried, were way better in this space than QGis.

In their (possible) defence, stepping outside the the narrow ESRI path of the way you're supposed to do things makes your life a massive pain. I used to work for a company that (among other things) developed tools for clients on top of ArcGIS Online. And I've had many calls where a client asks for a small and very reasonable change, and I embarrassingly have to explain that that is not the 'approved' way doing things in ArcGIS Online, so trying to make that change will essentially require us to rewrite a whole lot of components more or less from scratch and might no even be possible.

I've spent a large number of hours trying to fight ESRI software, lost more fights than I've won and even my victories have mostly been pyrrhic.

Thanks for the perspective, I'm entirely new to the space. We're supposed to use ArcGIS. So if this turns out to be the least painful path, I'll have to learn to deal with it...

ArcGIS the desktop tool, or ArcGIS Online? The desktop tool at least has a pretty complete (but pretty annoying) python API so that if/when you have to, you can circumvent the GUI and automate almost everything. ArcGIS Online is unfortunately extremely inflexible, and things that should be easy are often impossible.

Is there a better option for raw base tilesets? When making my app mapping world tides (https://solunar.pages.dev) I tried OSM for instance but it only had raster tiles at LoDPI resolutions available. ESRI on the other hand had full vector support in a variety of formats^ and provided great docs for integrating with FOSS rendering toolkits. The free tier seems generous, though once it runs out the price increases sharply.

^ For instance, a really neat one that renders ocean features in as much detail as the land typically gets

Well, the main other player here is Mapbox.

Mapbox's free tier is an order of magnitude less permissive. They even charge for loading their JavaScript: $5 per kiloload! Feels like the best approach if I'm not trying to commercialize is just to encourage people to self-host using their own API key.

Assuming the service ever even gets to 2M tile requests/month. That alone would be quite the accomplishment.

Though with Gaia going to hell, something new needs to come along.

Load osm data in postgres and generate the tiles yourself with something like geoserver/mapserver

The cost and effort of deploying something like that in production is decidedly non-trivial. I know because I've done it. Unless you need the customization and flexibility such a solution offers, it's rarely worth it.

If there is one thing I want AI to accomplish ;;

Have an AI crawl a RASTER dataset and VECTORIZE all lines into SVG.

I love QGIS. It's like a super useful Excel of mapping, and so much more.

I've had difficulty making it work for production services (especially extensions). Ultimately moved more towards the Python/OSM stack.

Such a fun space.

What made it problematic to run in prod for you?

Splines instead of points — boom! That sounds awesome.

Reply on news.ycombinator.com