WeRelate:Suggestions/Adopt a Flexible Place Page System like Gramps 4

Version 4 of the python open source genealogy program gramps has a very flexible hierarchical place model intended to be able to express the entire history of places as their hierarchy changes. Implementation details are that a place is represented by an opaque oid with its hierarchy all expressed as relationships. Then when displaying any event the event date is used to generate a time-dependent place title on the fly. (n.b. this means every event would need to specify a date or accept the default, which for consistency with present practice could be set to 1900.)

I do not envision this as something WR would adopt in the next couple of years, but if the gramps community demonstrates its usefulness perhaps the conversion could be done over a ten-year timeframe.

I'm not sure an opaque oid would work for WR, but perhaps each placename could be just a single level and have an integer disambiguation counter like we do for people.

--pkeegstra 20:59, 11 November 2015 (UTC)

We need to list the requirements that need to be solved, before listing a solution, i.e., gramps. Then perhaps a gramps-like solution will work, or perhaps not. But this makes a cost-benefit analysis possible.
One problem I see, to the limited extent I understand what gramps has, is that while there is a significant software effort to implement such a solution, there may be an even more significant data collection effort to build an accurate historical name database that drives it.
Another is that it still may not allow you to say what you mean. Take Salem, Massachusetts. Ignoring the name of the colony at the time, and the county it was in, parts of Salem became Beverly, Wenham and Danvers (Salem Village). What happens if I use the name Salem in, say, 1675. Do I mean Salem as that name existed in 1675 because I don't have a more precise knowledge of the location, or do I mean the part of old Salem that is still Salem today, a smaller, more precise specification of where an event happened? If you automatically assume the event's date is the key, you would use the larger region, but how would you enable to specify the smaller region?
In a sense, the 1900 name is like a name group, and it is possible to document all the history of that place on the Place page or with the links to external sources like wikipedia, etc. And in general (though maybe not always?), a name that existed in both 1600 and 1900 is going to have a smaller, more precise meaning in 1900 than 1600, so that when preciseness is available, i.e., when the father wills his homestead to the son, but the son is born in Salem and dies in Beverly, you know both events occurred in the part of Salem that is Beverly, you use Beverly even though no such town existed at the time of birth.
A related problem has to do with parishes and villages within a town. It is attractive to specify the most precise region possible when locating events, so why stop at 4 part names? --Jrich 00:05, 12 November 2015 (UTC)
We kicked this can around before, and it is a bad idea for the reasons that Jrich gives. We had a user some time ago who went around renaming the place on person pages in various colonial New England towns - so he would shorten, for example, Barnstaple, Massachusetts, United States to 'Barnstaple, Massachusetts' - when several of us objected to this, he then said we weren't being 'historically accurate'. But this user's 'historical accuracy' was selective, because 'Massachusetts' didn't exist at the time, and he failed to note that in the 1680s none of the old colonies technically existed, at least on paper, during the Dominion of New England period when the colonies were consolidated. Time-sensitive place pages would probably become a nightmare to sort because boundaries have proven flexible over the centuries. I realize that this is not the same as you call it, on the fly, but the maintenance on such a project alone (not to mention the confusingly fast changing political arrangements in some of these periods) make it much more trouble than it would ever be worth. Daniel Maxwell 07:56, 15 November 2015 (UTC)
JRich, the gramps answer to your specific question is that gramps (being a personal database) encourages very low-level places. So the farm you describe would be a place, and it would be contained in Salem initially and in Beverly after that is formed. I can't say if that would be the WR answer; the WR community would have to reach a consensus. (And for every place which is more expressive in 1900, I'm sure I can find you a counterexample where the 1900 structure fails to make distinctions which existed either earlier or later.)
(I would guess that the semantics is that the place definition represents the named place thru history at its extent at each time, in other words it has an implicit time-dependent polygon which expands and contracts as needed.)
FWIW, part of my motivation for proposing a way to get beyond the 1900 Rule is that some of the people working in what became Kentucky and the adjoining counties of Virginia have expressed to me that they find it difficult to express the full history of their places within the constraints of the 1900 Rule. (And other people just don't understand the 1900 Rule, since it's not easy to find specs for it except on Talk pages.)
DMaxwell, yes, populating the full history of a place from 800 to the present (in North America, first written records to the present) is a formidable task. The gramps community has set out to attempt that, and WR has a lot of info towards that in its place pages. And it will work like the present place pages; as local experts are found, the pages will be improved. And one of the things this will accomplish is to get beyond expressing colonial places by just hacking off "United States". One will be able to express all the shifts in colony structure, even the Dominion of New England. And once that's correct, all events in colonial New England will use those names. (So, I reject the idea that because something is being done badly under the present structure, it would b a "bad idea" to set up a structure so it could be done right.)
The priority thing I would like to ask the gramps group is how they would handle a place with split hierarchy, like a midwestern town spanning multiple counties. There are two cases both of which need to be able to be expressed: the case where you know just the city name, and the case where you know both the city and the county.
--pkeegstra 12:41, 15 November 2015 (UTC)
With respect to your last paragraph, I don't know anything about gramps, but that situation isn't very hard to deal with in a normal database where you can have one-to-many links. The name of the record in the database would be unique and not dependent on the name of any other entity recorded. The problem with the way the wiki names these places is that the page name is dependent on the names of multiple entities (ie. city, county, state, etc.). If the wiki named place pages in the same fashion as it names people (eg. Newton (1) [perhaps Newton, MA], Newton (2) [perhaps Newton, KS], etc.), you could have the usual one-to-many links without confusing people or causing disagreements. -Moverton 21:22, 16 November 2015 (UTC)
Good point. I missed that because I was looking at a different question. A good example is Holland, Michigan, which is in both Ottawa County and Allegan County. The existing WR system only allows one of those to be the place name, the other must be indicated as "also located in". (The choice is made the way it is because in 1900 Holland was still all on the Ottawa side of the county line.) The gramps setup allows both contained counties on an equal footing. The question I was addressing was different: sometimes one knows just e.g. that an event happened in the city Holland, but other times one has the extra information that it happened on the Ottawa County side or the Allegan County side. So it would be good to be able to handle both semantics. --pkeegstra 11:38, 23 November 2015 (UTC)

I am curious to know how the Gramps system has panned out now after a couple of years. I know that I can read all of the discussions myself, but I am most interested in a critique from someone who is familiar with our setup as well. I also want to share a link to the Atlas of Historical Boundaries project which allows for the download of their data for use in other projects and might be something useful. --cos1776 14:12, 28 May 2017 (UTC)

Some people love it, some people find it too flexible. To facilitate setting up the data, they have a tool which pulls appropriately formatted data from the GOV database developed by the Verein für Computergenealogie. Myself, I must confess I haven't upgraded yet because my system doesn't have all of the prerequisites for Python 3. --pkeegstra 21:20, 28 May 2017 (UTC)
Another update - randymajors is also working on this idea at http://www.randymajors.com/p/maps.html . --cos1776 18:05, 16 July 2017 (UTC)