I know the boat sailed in early October but still felt compelled to respond given I have an interest in both disciplines. I saved Thierry Gregorius' blog post to Pocket and finally read it last week. In general I get the sentiment and frustration but found myself disagreeing with a number of points, which is good because I wouldn't want him retiring from blogging either! So this is less of a critique and meant more to initiate (instigate?) more dialogue - Thierry, you did ask for more comments.

First I disagree that the field of GIS itself is less simple now - it was never simple. I'm referring to the underlying science related to map projections, re-projections, neighborhood statistics, autocorrelation - pick your topic. Tobler's 1970 seminal paper on autocorrelation is instructive - it was never simple. If you consider the GIS software itself, the GUIs have certainly become more cluttered but in every package I have used it was possible to turn off toolbars and extensions/plugins. If this is really a concern, vilifying software and vendors should cut across all sectors; why is no one clamoring for Cricket Graph, a return to Fortran or Internet Explorer? Ok, that last one wasn't nice. I completely agree with the final conclusion that 'spatial is not special' but then again, neither are geoscientists (I just implied I wasn't special too, don't get feisty).

Geoscientists are not the only folks who consume data that has been 'digitized from analog sources, scraped from literature or intelligence reports, imported from spreadsheets, dug up from archives, copied from network drives, extracted from databases, downloaded from web feeds" or have to develop a story using a variety of data formats and datasets (Gregorius, 2013). The same holds true for town planners (not necessarily GIS professionals), historians, biologists, transportation specialists, and many (many) other disciplines. 'Our' issue is not ours alone to claim. 

The data format issue is real and it is challenging but certainly not insurmountable. I find it unrealistic to assume that if geoscientists "don't consider themselves to be GIS professionals" but that "no GIS professional today can survive without FME or GDAL" that somehow the vendors have ignored the needs of our profession. Is this issue related to vendors producing software capable of consuming the data or that many data sources are stored in proprietary data formats with complex schemas that require translation? FME and GDAL are completely different in my mind: FME offers translation capabilities in addition to conversion while GDAL supports raster conversion and manipulation (OGR is the vector equivalent). GDAL and OGR are a means to address the issue of data formats not complicate it - both serve us well in the sea of data formats. So, vendors nor software 'require' us to create silos, we (yes, geoscientists too) choose that route (link, link). Criticizing open standards or open-source GIS packages is akin to couples bemoaning how challenging the adoption process is... there are plenty of children in foster care, they just aren't adopted.

The GeoWeb can be abused and neglected by any profession - geologic clearinghouses  commit some of the more egregious sins, which result in slow and unfriendly user interfaces (I won't point to at any particular state survey here, but do a random sampling on your own). Eric Edmonds and Brian Timony have offered thoughts on why these portals are so cumbersome but that isn't because of the GIS industry or GIS professionals per se. Similarly, I'm not sure how developers creating mobile applications - who aren't 'GIS people' or geologists - would know what geoscientists need so they didn't feel like they fell through the cracks. The implication here is that we need apps that can perform 'geological data interpretation' in the field on a smartphone or tablet versus having the capability to collect points, lines and polygons and easily enter data using something like a data dictionary (those apps exist).

Open-source GIS does provide a viable alternative to a proprietary vendor marriage (this isn't 'new' - GRASS has offered this for 30-years). The various FOSS-GEO options are not the reason we struggle with proprietary formats, again it is the people who choose store them that way. And Google Earth certainly offers an easier interface to work in but I don't think Brian claimed that 'most enterprise GIS requirements would be met' but instead referred to 'data viz' needs. It's a semantic issue but in my mind Google Earth is not a GIS. You can create, store, and visualize spatial data but is missing the most fundamental (in my mind) tools: the ability to query and analyze spatial relationships. So if you want to create a visual depiction of your data, store it in network KMLs and distribute, Timoney's right (again!). But if you want to buffer private water wells, use zonal statistics to characterize the dominant bedrock within a watershed, interpolate an isopach map, etc... you can't do that using Google Earth. If you're 'really' nostalgic for ArcView 3.x then take a look at gvSIG, it has a very similar interface and functionality.

I'm a little unclear who the GIS behind the curtain actually IS but here are some thoughts on Thierry's questions:
  1. Where are the innovative solutions for dealing with the variety of geoscience data?

    Timoney referred to this in one of his Tweets (circa 2007): Store your data in a PostGIS enabled PostgresSQL database and become a spatial data hipster.

  2. Where are the productivity tools to simply assemble digital scrapbooks of georeferenced information?

    As far as I can tell, QGIS isn't planning a coup of ESRI (nor does their mission allude to that), the interface is entirely customizable, it consumes data stored in PostGIS AND it has the ability to create map books using the Atlas function.

  3. Where are the flexible data models that enable thematic harvesting and analysis irrespective of data type?

    PostGIS via command line, the SPIT and DB Manager plugins for QQIS or ST-Links Spatial Links for ESRI.

  4. Where are the analytical tools that can handle dirty and incomplete data without hours of pre-processing?

    If it is dirty, it will probably take some tinkering... I guess I'd have to see an example but my default would be Open Refine. Or just import it into a PostGIS database...
  5. Where are the predictive user interfaces that only show relevant options?

    Personally I don't want a software package deciding what is relevant to me, I'd much rather selectively turn off things I don't need and set a default workspace. I'm more frustrated by this in Pages or Excel than any GIS. By the way, check out PostGIS.

So I don't think the GIS community has ignored or let geoscientists fall through the cracks and I certainly don't think geologic data is special or any messier than other disciplines. If you're looking for a geology-flavored GIS, check out BeeGIS it's built on JGrass so you get the spatial tools of GRASS in a simple interface that has tools specific to hydrology and geomorphology. And that's the ultimate rub for me in this discussion - if the existing GIS packages aren't sufficient, perhaps the geoscience community should follow Matt Hall's example and build their own. Waiting for a vendor or non-geoscientist to create a custom tailored, predictive GIS will likely happen sometime in 20... never. BUT, I encourage you to explore PostGIS and let Pele change your world - she's cute, she's talented and she'll eat up all your spatial data.